See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224381355

Low-Temperature Two-Phase Microchannel Cooling for High-Heat-Flux

Thermal Management of Defense Electronics

Article  in  IEEE Transactions on Components and Packaging Technologies · July 2009

DOI: 10.1109/TCAPT.2008.2005783 · Source: IEEE Xplore

CITATIONS READS
69 209

2 authors, including:

Issam Mudawar
Purdue University
258 PUBLICATIONS   12,650 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Progress of Boiling Two-Phase Flow and Heat Transfer View project

Flow boiling in micro-channel heat sink View project

All content following this page was uploaded by Issam Mudawar on 27 May 2014.

The user has requested enhancement of the downloaded file.

IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 32, NO. 2, JUNE 2009 453

Low-Temperature Two-Phase Microchannel Cooling

for High-Heat-Flux Thermal Management
of Defense Electronics
Jaeseon Lee and Issam Mudawar

Abstract—For a given heat sink thermal resistance and ambient Hydraulic diameter.
temperature, the temperature of an electronic device rises fairly Mass velocity.
linearly with increasing device heat flux. This relationship is es-
pecially problematic for defense electronics, where heat dissipa- Heat transfer coefficient.
tion is projected to exceed 1000 W cm2 in the near future. Di- Microchannel height.
rect and indirect low-temperature refrigeration cooling facilitate
appreciable reduction in the temperature of both coolant and de- Latent heat of vaporization.
vice. This paper explores the benefits of cooling the device using di- Distance between thermocouple and base of
rect and indirect refrigeration cooling systems. In the direct cooling
system, a microchannel heat sink serves as an evaporator in a con- microchannel.
ventional vapor compression cycle using R134a as working fluid. Modified Jacob number.
In the indirect cooling system, HFE 7100 is used to cool the heat Thermal conductivity.
sink in a primary pumped liquid loop that rejects heat to a sec-
ondary refrigeration loop. Two drastically different flow behaviors Length of microchannel.
are observed in these systems. Because of compressor performance Fin parameter.
constraints, mostly high void fraction two-phase patterns are en-
countered in the R134a system, dominated by saturated boiling. Total number of microchannels in heat sink.
On the other hand, the indirect refrigeration cooling system fa- Nusselt number.
cilitates highly subcooled boiling inside the heat sink. Both sys-
tems are shown to provide important cooling benefits, but the in- Pressure.
direct cooling system is far more effective at dissipating high heat Pressure drop across micro-channel.
fluxes. Tests with this system yielded cooling heat fluxes as high
as 840 W cm2 without incurring critical heat flux (CHF). Results Prandtl number.
from both systems are combined to construct an overall map of Heat flux through heat sink base area.
performance trends relative to mass velocity, subcooling, pressure,
and surface tension. Extreme conditions of near-saturated flow, Mean heat flux through three conducting walls of
low mass velocity, and low pressure produce “micro” behavior, microchannel.
where macrochannel flow pattern maps simply fail to apply, insta- Reynolds number.
bilities are prominent, and CHF is quite low. One the other hand,
systems with high mass velocity, high subcooling, and high pressure
Thermal resistance.
are far more stable and yield very high CHF values; two-phase flow Temperature.
in these systems follows the fluid flow and heat transfer behavior
as well as the flow pattern maps of macrochannels.
Local bulk liquid temperature.
Index Terms—Electronics cooling, high heat flux, microchannel
Measured copper temperature.
flow, phase change. Local base–wall temperature of microchannel.
Microchannel width.
NOMENCLATURE Weber number.
Aspect ratio of microchannel. Modified Weber number.
Boiling number. Half-width of wall separating microchannels.
Specific heat. Martinelli parameter.
Thermodynamic equilibrium quality.
Stream-wise coordinate.
Manuscript received April 03, 2008; revised July 15, 2008. First published
February 06, 2009; current version published July 22, 2009. This work was sup- Greek Symbols:
ported by the Office of Naval Research (ONR). This work was recommended
for publication by Associate Editor R. Prasher upon evaluation of the reviewers
comments. , Channel aspect ratio ( ).
J. Lee is with the United Technologies Research Center (UTRC), E. Hartford,
CT 06108 USA. Fin efficiency.
I. Mudawar is with the Boiling and Two-Phase Flow Laboratory (BTPFL)
and the Purdue University International Electronic Cooling Alliance (PUIECA), Viscosity.
West Lafayette, IN 47907-2088 USA (e-mail: [email protected]). Density.
Digital Object Identifier 10.1109/TCAPT.2008.2005783

1521-3331/$25.00 © 2009 IEEE

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
454 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 32, NO. 2, JUNE 2009

Surface tension. the device and coolant and fixed coolant temperature, the
Fluid phase. device temperature increases fairly linearly with increasing
heat dissipation rate. This relationship could easily bring the
Subscripts: temperature of the device above its maximum limit when
3 Based on three-sided heating of rectangular dissipating high heat fluxes. For a device operating slightly
microchannel. below its maximum temperature limit, dissipating increasing
Half-way along microchannel. amounts of heat requires reducing the temperature of the
Microchannel. liquid coolant. This trend would ultimately drop the coolant
Thermodynamic equilibrium temperature below ambient temperature, requiring the use of
a refrigeration system to sustain the cooling. Further reducing
Liquid. the coolant temperature allows the device to operate well below
Liquid only. its maximum temperature limit. Various refrigeration cooling
Vapor. schemes have been suggested to achieve this objective [2], [3].
Microchannel inlet. Of those, vapor compression systems have attracted the most
attention because of their relatively high cooling capacity at
Microchannel outlet. temperature of interest [4]–[7].
Solid. Refrigeration cooling can be implemented in two different
Saturation. configurations. In the first direct-refrigeration-cooling config-
Single-phase. uration, the cooling module is incorporated as an evaporator in
a vapor compression cycle, and the refrigerant serves as coolant
Thermocouple. for the electronic device. The alternative indirect-refrigeration-
Two-phase. cooling configuration involves using two fluid loops. Heat from
Channel base wall. the device is rejected to a primary coolant circulating through a
pumped liquid loop that rejects the heat via a heat exchanger to
I. INTRODUCTION refrigerant flowing in a separate vapor compression cycle.
In recent years, microchannel cooling has attracted the
A. Application of Refrigeration Cooling most attention of any cooling scheme for high-flux electronic
cooling. The small hydraulic diameter of a microchannel heat
HE PAST 30 years have witnessed unprecedented en- sink greatly increases the convective heat transfer coefficient,
T hancement in chip performance, brought about mostly
by advances in microminiaturization of electronic circuits. As
especially where the coolant is allowed to vaporize along
the channel. Microchannel heat sinks also possess attributes
the number of circuits integrated in a single device increased, that are of special importance to electronics cooling such as
so did the amount of dissipated heat. This problem is further compactness and minimal coolant usage. Allowing the coolant
compounded when multiple devices are packaged in close to vaporize along the microchannel contributes the additional
proximity to one another on a single circuit board, especially in benefits of significant enhancement in the heat transfer coeffi-
systems containing a multitude of circuit boards. cient and better axial fluid and surface temperature uniformity,
During the 1980s, device heat dissipation was easily man- dictated mostly by the coolant’s saturation temperature, com-
aged using air-cooled heat sinks, and any increases in device pared to single-phase microchannel flow. However, two-phase
heat flux could be tackled with larger and more effective heat microchannel heat sinks are not without shortcomings. Small
sink designs. It was not until the early 1990s, as device heat hydraulic diameter can lead to appreciable pressure drop and
flux approached 100 W cm , that the electronics industry was therefore high pumping power consumption. Two-phase mi-
forced to begin exploring liquid cooling solutions. Recent liquid crochannel heat sinks are also prone to different types of flow,
cooling research efforts have culminated in numerous cooling pressure, and temperature oscillations [8]. The cooling perfor-
solutions that are capable of tackling mostly the 50–150 W cm mance of a heat sink depends on whether it is used in a direct-
range [1]. or indirect-refrigeration-cooling configuration. As discussed
Defense electronics represent a specialized class of devices below, these configurations yield two drastically different types
that have for the most part followed the heat dissipation trends of flow boiling, saturated and subcooled.
of commercial devices. However, a new generation of devices
for defense radars and directed-energy laser and microwave B. Direct Refrigeration Cooling and Saturated Microchannel
weapons are approaching 1000 W cm [1], which exceeds the Flow Boiling
capabilities of today’s most advanced liquid cooling solutions. Using a microchannel heat sink as an evaporator in a di-
This trend is the primary motivation for the present study. rect-refrigeration cooling system requires that the coolant’s
The goal here is to develop advanced thermal management operating conditions conform to those of a vapor compression
schemes that can remove very large heat fluxes from advanced cycle. First, a refrigerant (R134a for most modern systems)
defense electronics while maintaining device temperatures is used as the primary coolant. The refrigerant enters the
below 125 C. microchannel evaporator as a two-phase mixture and exits as
The difficulty in implementing even the most aggressive saturated or superheated vapor as required by most refrigera-
liquid cooling scheme is that, for a given resistance between tion compressors. Under these conditions, as well as in most

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
LEE AND MUDAWAR: LOW-TEMPERATURE TWO-PHASE MICROCHANNEL COOLING 455

published microchannel studies, the coolant undergoes rapid renders the task of achieving very high heat fluxes while main-
change of phase into a high void fraction mixture. Therefore, taining the device temperature below 125 C quite elusive.
slug and annular flow are often cited as dominant flow regimes This paper explores the cooling potential of two-phase
for this cooling configuration. Hence, two-phase pressure drop microchannel heat sinks when combined with refrigeration
and heat transfer coefficient models are based mostly on these cooling. Both direct and indirect refrigeration cooling are con-
regimes [9]–[21]. sidered, and dominant two-phase flow patterns explored. The
The authors of the present study examined the performance of pressure-drop and heat transfer performances are discussed for
a microchannel heat sink that served as an evaporator in a direct each configuration. These and other practical considerations
refrigeration cooling cycle using R134a as working fluid [20], are used to identify the merits and drawbacks of each config-
[21]. Aside from developing new correlations for two-phase uration, especially the ability to meet the high-flux cooling
pressure drop and heat transfer coefficient, they provided prac- requirements of defense electronics.
tical insight into the implementation of the direct-refrigeration-
cooling configuration. They showed that the throttling valve of- II. EXPERIMENTAL METHODS
fers significant stiffening to the refrigeration cycle, suppressing
the large pressure oscillations common to microchannel heat A. Direct Refrigeration Cooling System
sinks.
Fig. 1(a) shows a schematic diagram of the test loop used
C. Indirect Refrigeration Cooling and Subcooled to simulate a direct-refrigeration-cooling system. The mi-
Microchannel Flow Boiling crochannel test module is incorporated as evaporator in a fairly
Using a separate loop for the primary coolant provides conventional vapor compression cycle using R134a as working
greater flexibility in attaining the desired microchannel heat fluid. The loop utilizes a rotary compressor powered by a dc
sink’s inlet conditions. Most importantly, the coolant does not source and a finned-tube air-cooled condenser. The condenser’s
have to be maintained in a near-saturated or superheated state as air flow is regulated by a variac to control the condenser’s
required by the compressor in the direct-refrigeration-cooling exit subcooling. A glass flow meter is used to confirm liquid
configuration. state at the condenser outlet as well as measure volumetric
Subcooled boiling occurs when the coolant is supplied to the flow rate. A manual metering valve situated upstream of the
heat sink below saturation temperature, corresponding to a ther- microchannel evaporator throttles the flow between the high
modynamic equilibrium quality below zero. The quality rises condenser pressure and low evaporator pressure.
along the microchannel and, if the channel is long enough, the Fig. 1(b) shows the construction of the test module
flow may ultimately transition to saturated boiling. However, the used in the direct-refrigeration-cooling system. Fifty-three
short length of a microchannel heat sink precludes such transi- 231- m-wide and 713- m-tall microchannels are cut into the
tion, especially for high mass velocities and highly subcooled top 2.53-cm-long surface of an oxygen-free copper block. This
inlet conditions. Because of large differences in void fraction, block is inserted into an insulating fiberglass plastic housing
the heat transfer mechanism for subcooled boiling is categor- and covered atop with a transparent plate of polycarbonate
ically different from that for saturated boiling. In subcooled plastic. Heat is provided by three cartridge heaters inserted
boiling, liquid flow is more abundant and phase-change occurs into the lower enlarged portion of the copper block. Coolant
mostly by bubble formation at the wall. In general, much higher temperature and pressure are measured in the housing plenums
critical heat flux (CHF) values are possible with subcooled flow both upstream and downstream of the microchannels. The
boiling than with saturated. Furthermore, CHF mechanism for copper block temperature is measured by a T-type thermo-
subcooled boiling is fundamentally different from that for satu- couple inserted beneath the microchannels. Table II provides
rated boiling. Subcooled boiling CHF (also referred to as “de- details of the microchannel geometry.
parture from nucleate boiling”) is the result of localized vapor Operating conditions for the microchannel in the direct-re-
blanket formation along the heated wall, and occurs despite the frigeration-cooling system are as follows: inlet pressure of
abundance of liquid in the core. On the other hand, CHF in sat- bar, inlet quality of ,
urated boiling occurs in the channel’s downstream liquid defi- outlet quality of —superheat, and mass velocity
cient region as a result of dry-out of the annular liquid film. of kg m s. Further details of this system are
Research on subcooled flow boiling in microchannels pre- available elsewhere [20], [21].
dates microchannel electronic cooling studies. This form of
B. Indirect Refrigeration Cooling System
cooling was the subject of intense study since the mid 1970s at
the Massachusetts Institute of Technology Energy Laboratory Fig. 2(a) shows the flow diagram of the indirect-refrigera-
for cooling of electrodes in magnetohydrodynamic energy con- tion-cooling system. Here, the vapor compression system is
verters and turbine blades [22]. A number of subsequent studies completely isolated from the primary cooling loop containing
demonstrated the enormous cooling potential of subcooled flow the microchannel module. The working fluid in the primary
boiling. Mudawar and Bowers [23] showed highly subcooled cooling loop is 3M’s Novec fluid HFE 7100. This fluid has
and high mass velocity flow boiling of water in small diameter a low freezing point below 100 C and a moderate boiling
tubes could safely dissipate up to 27 000 W cm . point of 60 C at atmospheric pressure. Table I compares
Unfortunately, the dielectric coolants recommended for elec- representative values of the thermophysical properties of HFE
tronics cooling possesses poor thermal transport properties. This 7100 to those of R134a.

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
456 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 32, NO. 2, JUNE 2009

Fig. 1. (a) Schematic diagram of flow loop and (b) construction of test module for direct refrigeration cooling system.

TABLE I
SUMMARY OF SATURATED THERMOPHYSICAL PROPERTIES OF R134a AND HFE 7100 AT 1.0 BAR

TABLE II the indirect cooling microchannel test module. Further details

MICROCHANNEL TEST MODULE DIMENSIONS of this test module are available elsewhere [24], [25].
The indirect-refrigeration-cooling experiments were per-
formed at two test module inlet temperatures, 30 and 0 C,
at a constant outlet pressure of 1.138 bar. Mass velocity covered
a range of kg m s.

III. PRESSURE DROP RESULTS

As indicated earlier, high pressure drop is a primary concern
As shown in Fig. 2(a), HFE 7100 liquid is pumped through in the implementation of microchannel cooling. One reason for
a heat exchanger to reduce its temperature by rejecting heat to the high pressure drop is the small hydraulic diameter that is
the secondary refrigeration loop. Exiting the heat exchanger, the required to achieve high convective heat transfer coefficients.
HFE 7100 liquid passes through a filter followed by a Coriolis Another reason is vapor production along the microchannels,
mass flow meter before entering the microchannel test module. which increases both the frictional and accelerational pressure
Throttling valves situated upstream and downstream of the test gradients along the microchannel as well as overall pressure
module are used to regulate both flow rate and the test module’s drop. A third reason is that high heat fluxes require high mass ve-
outlet pressure. locities, which also contribute high pressure drop. However, the
Fig. 2(b) illustrates the construction of the microchannel test role and significance of each of these parameters differs greatly
module. This module is fairly similar to the one used in the between saturated and subcooled flows.
direct-refrigeration-cooling system, save base area and number When comparing saturated and subcooled flows, it is useful to
of cartridge heaters. Table II provides the detailed geometry of examine pressure drop per unit microchannel length, especially

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
LEE AND MUDAWAR: LOW-TEMPERATURE TWO-PHASE MICROCHANNEL COOLING 457

Fig. 2. (a) Schematic diagram of flow loop and (b) construction of test module for indirect refrigeration cooling system.

for different microchannel lengths. Such experimental data are thermophysical properties, the high-pressure drop of saturated
presented in Fig. 3. To calculate the microchannel pressure drop, boiling R134a is the result of high void fraction. In contrast,
the inlet plenum pressure loss and outlet plenum pressure re- the subcooled boiling HFE 7100 system maintains single-phase
covery are subtracted from the total pressure drop measured by liquid flow and/or minimal void fraction over the noted heat flux
pressure transducers connected to the plenums. Details of the re- range.
lations used to account for these effects are available elsewhere The second noteworthy difference between Fig. 3(a) and (b)
[8], [20]. is overall shape of the pressure drop characteristics. A key factor
More than attempting to directly compare results for R134a that influences this shape is the onset of boiling at a particular
direct refrigeration cooling and HFE 7100 indirect cooling, heat flux value. Thereafter, pressure drop begins to increase ap-
Fig. 3 is intended to identify trends in the variation of pressure preciably due the strong influence of void fraction on pressure
drop per unit length as well as gain a better understanding of the drop. Mass velocity can complicate the pressure drop trends for
underlying physical mechanisms. It should also be emphasized certain conditions.
that significantly higher mass velocities are possible with the The R134a data in Fig. 3(a) correspond exclusively to satu-
indirect-refrigeration-cooling system, given the greater flexi- rated boiling conditions at the microchannel inlet brought about
bility in increasing coolant flow rate with this system compared by flashing across the throttling valve. Saturated or superheated
to the direct-refrigeration-cooling system. The higher pressure conditions are maintained along the length of the microchannel.
drop values in Fig. 3(b) can be explained by the pressure For each mass velocity, Fig. 3(a) shows pressure drop rises more
drop being proportional to the square of mass velocity. Notice rapidly at low heat fluxes than at high fluxes. This trend may be
that any differences in pressure drop due to microchannel explained by downstream microchannel dry-out at high fluxes
geometry are minimized by comparing data for similarly sized versus predominantly slug and annular flow along the entire mi-
microchannels. crochannel at low fluxes. The extent of the downstream dry-out
Aside from the pressure drop trends, a noteworthy compar- region increases with increasing heat flux, reducing the frac-
ison of Fig. 3(a) and (b) is the magnitude of pressure drop per tion of the length undergoing high pressure saturated boiling.
unit length for the lowest mass velocity case for HFE 7100 Fig. 3(a) shows a predictable monotonic trend of increasing
versus the highest mass velocity case for R134a for the same pressure drop with increasing mass velocity.
heat flux range of W cm . Despite a nearly 2.8 The HFE 7100 pressure drop data in Fig. 3(b) show far more
times greater mass velocity for the former, pressure drop is much complicated roles for heat flux and mass velocity. Notice how
smaller than for R134a. Aside from the obvious differences in pressure drop for each mass velocity first decreases slightly with

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
458 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 32, NO. 2, JUNE 2009

along the middle section of the microchannel corresponding to

the most frequently observed high void patterns of elongated
slug flow and annular flow for R134a. On the other hand,
Fig. 4(b) depicts the bubbly flow pattern that dominated the
HFE 7100 subcooled boiling tests.
Additional extensive flow visualization results are available
in previous studies by the present authors for both saturated
boiling [20] and subcooled boiling [25]. These studies provide
comprehensive assessments of prior pressure drop correlations
and recommend new correlations with greater accuracy in pre-
dicting experimental data.

IV. HEAT TRANSFER RESULTS

A. Reduction of Heat Transfer Coefficient Data

Fig. 5 shows a unit cell consisting of a single microchannel,
surrounding copper half-walls, and portion of the top cover
plate. A simplified fin model of the copper walls is used to
construct an energy balance for this unit cell. Equating the heat
influx to the unit cell through the bottom solid wall to the heat
efflux by flow boiling along the microchannel bottom wall and
sidewalls yields

(1)
where is the fin efficiency. Since the top wall is adiabatic, the
fin efficiency is given by [26]

(2)

where is the fin parameter defined as

(3)
Fig. 3. Variation of pressure drop per unit microchannel length with heat flux
for (a) direct refrigeration cooling system and (b) indirect refrigeration cooling The fin base temperature is calculated using the assump-
system. tion of one-dimensional heat diffusion between the plane of the
thermocouple embedded in the copper block and the plane con-
increasing heat flux. This low heat flux region corresponds to taining the channel base
single-phase liquid flow, and the decreasing pressure drop is the
result of a reduction of liquid viscosity with increasing temper- (4)
ature. This trend is reversed once nucleate boiling commences
inside the microchannel, resulting in a minimum in the pressure The fluid temperature in (1) is the fluid bulk tempera-
drop dependence on heat flux. It should be emphasized that all ture halfway along the microchannel. This temperature is de-
data in Fig. 3(b) correspond to subcooled boiling conditions, termined differently for the R134a and HFE 7100 systems. For
where bubbles are generated in a thin superheated liquid layer the R134a saturated boiling tests is based on pressure at the
at the wall while the bulk liquid core is subcooled. As bubbles same location determined from the measured inlet pressure and
detach from the wall, they have a tendency to condense in the pressure drop correlation scheme presented in [21]. For the HFE
liquid core, which explains the relatively low void fractions en- 7100 subcooled boiling tests, is assumed equal to the average
countered in subcooled boiling. The milder increase in pressure of the measured inlet and outlet temperatures of the subcooled
drop with increasing heat flux at kg m s compared flow.
to the lower mass velocities is the result of stronger condensa- Equations (1) and (2) are used to determine since all other
tion effects at higher mass velocities. parameters are easily measured or calculated.
Fig. 4 shows representative images for saturated R134a and
subcooled HFE 7100 flows that confirm the aforementioned B. Heat Transfer Coefficient Results
influence of void fraction in the pressure drop trends. Saturated Differences in system operation caused the R134a and HFE
boiling spans most two-phase flow regimes, bubbly, slug, 7100 tests to be performed in different ways. Since the com-
elongated slug, annular, and dry out, though bubbly flow is only pressor in the direct refrigeration cooling system is sensitive to
observed at very low heat fluxes. Fig. 4(a) captures conditions changes in heat load, varying power input to the test module (to

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
LEE AND MUDAWAR: LOW-TEMPERATURE TWO-PHASE MICROCHANNEL COOLING 459

Fig. 4. Microchannel flow boiling pattern in central region of microchannels for (a) R134a saturated boiling (direct refrigeration cooling system) at
G = 163 =kg m 1 s, x = 0:53 q = 31:6 =
, and W cm , and (b) HFE 7100 subcooled boiling (indirect refrigeration cooling system) at G = 1; 339 =
kg m 1 s,
x = 0:570 , and q = 102:3 =
W cm .

as decoupling flow rate from heat input. These experiments were

performed by increasing heat flux while fixing flow rate.
Fig. 6 shows local heat transfer coefficient data from each test
section halfway along the microchannel plotted against thermo-
dynamic equilibrium quality, which is defined as

(5)

Notice that values in Fig. 6 are positive for all the R134a
saturated boiling data and negative for the HFE 7100 subcooled
boiling data.
Fig. 6(a) shows R134a saturated boiling produces very high
heat transfer coefficient values for the low range corre-
sponding to mostly bubbly flow. However, these values drop
sharply with increasing quality in the annular regime and ulti-
mately converge into heat transfer coefficient values for pure
vapor flow. As discussed in [21], the high heat transfer coeffi-
cients at low quality are unique to low surface tension fluids but
far less prevalent for high surface tension fluids. Low surface
tension fluids produce relatively small bubbles, allowing the
bubbly and slug flow regimes to persist for certain operating
condition. However, nucleating bubbles in high surface tension
Fig. 5. Unit cell used to infer heat transfer coefficient in microchannel. fluids such as water are much larger, tending to quickly fill
the channel and cause abrupt transition into annular flow. This
phenomenon allows R134a to yield heat transfer coefficient
alter heat flux) causes changes in flow rate as well as all key values comparable to those of water, despite the much poorer
flow parameters of the test module. Therefore, the R134a tests thermophysical properties of R134a. However, this attribute
were performed by fixing the heat input and varying the flow of R134a is compromised by the system requirements of the
rate with the aid of the throttling valve. The HFE 7100 system vapor compression cycle. On one hand, Fig. 6(a) proves quality
provided far greater flexibility in controlling parameters as well must be kept low to achieve high values. On the other hand,

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
460 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 32, NO. 2, JUNE 2009

Fig. 6. Variation of measured local heat transfer coefficient with equilibrium

quality for different heat fluxes and mass velocities for (a) R134a saturated
boiling in direct refrigeration cooling system and (b) HFE 7100 subcooled
boiling in indirect refrigeration cooling system.

Fig. 7. Boiling curves for (a) R134a saturated boiling in direct refrigeration
cooling system and (b) HFE 7100 subcooled boiling in indirect refrigeration
the compressor favors saturated or superheated vapor condi- cooling system.
tions. This requires the use of a secondary heater or phase
separator downstream of the test module, which compromises
the coefficient of performance of the vapor compression cycle. ability of the copper block to sustain downstream dry-out by
Another drawback of the R134a direct refrigeration system axial conduction.
is the relatively large wall temperature rise along the flow Fig. 7(b) shows boiling curves for HFE 7100. The ability of
direction, which is caused by rapid succession of different flow the indirect refrigeration cooling system to achieve high mass
patterns along the microchannel [27]. velocities allowed these tests to dissipate very high fluxes.
Fig. 6(b) shows the heat transfer coefficient for HFE 7100 Notice the highest mass velocity condition facilitating the
is constant for very low values corresponding to pure liquid dissipation of over 840 W cm , even though this particular
flow. The heat transfer coefficient begins to increase once nu- test was terminated before CHF in order to prevent excessive
cleation commences as subcooled boiling is initiated along the overheating of the test module components. Overall, increasing
microchannel. mass velocity increased both the single-phase heat transfer
coefficient and CHF; however, data appear to converge in the
C. Boiling Curve and Critical Heat Flux Trends nucleate boiling region.
Fig. 7 shows boiling curves for R134a and HFE 7100. The
D. Thermal Resistance Trends
absence of single-phase liquid data for R134a in Fig. 7(a) is the
result of the aforementioned compressor requirement. Another As indicated earlier, it is difficult to compare the perfor-
feature of the R134a data is the difficulty identifying CHF. One mances of the direct and indirect refrigeration systems based
reason for this difficulty is the relatively small heat flux range on heat transfer coefficient values. However, somewhat similar
of the direct refrigeration cooling tests. Another reason is the solid wall temperatures were maintained in both systems for a

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
LEE AND MUDAWAR: LOW-TEMPERATURE TWO-PHASE MICROCHANNEL COOLING 461

TABLE III
HEAT TRANSFER COEFFICIENT CORRELATIONS

E. Overall Assessment of Direct Refrigeration and Indirect

Refrigeration Systems
To provide a more quantitative comparison of the direct
refrigeration and indirect refrigeration systems, previous corre-
lations developed by the authors are used. Table III summarizes
microchannel heat transfer correlations for saturated boiling
[21] and subcooled boiling [25]. For a more systematic com-
parison, the same geometry is used and same mean heat flux,
W cm , is assumed for the three metal surfaces of
the microchannel. For the R134a saturated boiling case, quality
varies from 0.1 at the microchannel inlet to 1.0 at the outlet.
The HFE 7100 case involves subcooled boiling with an inlet
Fig. 8. Convective thermal resistance versus heat flux for (a) R134a saturated quality of and an outlet quality . The mass velocity
boiling in direct refrigeration cooling system and (b) HFE 7100 subcooled required to achieve these conditions is determined from the
boiling in indirect refrigeration cooling system.
following equations.

Saturated boiling
given value of superheat. This provided a basis for comparing
(7)
the thermal performances of the two systems based on thermal
resistance. The convective thermal resistance is defined from a Subcooled boiling
chip cooling perspective (rather than simply the reciprocal of (8)
) as
Table IV lists all conditions used in this comparative analysis.
Fig. 9 shows the results of the comparative study. Shown are
(6) variations of the two-phase heat transfer coefficient and the sur-
face-to-fluid temperature difference along the microchannel; the
Fig. 8 compares the resistance characteristics of the two sys- surface temperature used here is the base wall temperature of
tems. The R134a system is capable of resistances as low as 0.25 the microchannel. Fig. 9(a) show the results for R134a saturated
K cm W; however, these resistance values are maintainable boiling. Discontinuities in the two curves in this figure are the re-
up to only 100 W cm because of the dry-out effects. On the sult of using different correlations corresponding to low quality
other hand, the indirect cooling system facilitates much broader versus high quality. As discussed earlier, the heat transfer coef-
ranges of both mass flux and CHF, and thermal resistance values ficient is very high near the inlet, but decreases sharply towards
drop sharply once subcooled boiling commences along the mi- the outlet. Correspondingly, the wall temperature is lowest near
crochannels. The resistance for HFE 7100 decreases monoton- the inlet but increases appreciably at the outlet. Fluid temper-
ically with increasing heat flux, culminating in values around ature changes very slightly along the microchannel since it is
0.10 K cm W. equal to the saturation temperature.

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
462 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 32, NO. 2, JUNE 2009

TABLE IV
CONDITIONS USED IN COMPARATIVE STUDY (P = 1 bar, q = 200 W=cm )

Fig. 10. Flow regimes, heat transfer regimes, and variations of wall temperature
and convective heat transfer coefficient along microchannel flow.

assumes a fairly long heated channel. With a subcooled inlet,

Fig. 9. Predictions of heat transfer coefficient versus thermodynamic equi- liquid flow persists for a finite length until boiling commences
librium quality for (a) R134a saturated boiling in direct refrigeration cooling along the wall, marking the transition to subcooled boiling. No-
system and (b) HFE 7100 subcooled boiling in indirect refrigeration cooling
system. tice that the subcooled boiling region is comprised of two sub-
regions. The upstream highly subcooled region precludes any
appreciable bubble growth or entrainment of bubbles into the
Fig. 9(b) shows the predictions for the HFE 7100 subcooled liquid core. As more of the wall liquid is heated above the satu-
boiling case. Here, the heat transfer coefficient follows a trend ration temperature, condensation along the bubble interface be-
opposite to that of R134a. The heat transfer coefficient increases comes weaker and bubbles are able to grow larger as well as
and wall-to-fluid temperature difference decreases along the mi- depart into the liquid core. The transition point between the two
crochannel. Because the fluid temperature also increases along subcooled subregions is termed the “point of net vapor gener-
the microchannel, the wall temperature does not increase as it ation.” The saturated boiling region, which consists of bubbly,
does with R134a, and may therefore be reduced by reducing slug, and annular flow commences at the axial location were
the fluid inlet temperature. This is a key advantage of the indi- . This annular flow region persists until the point of
rect-refrigeration-cooling system. wall dry-out. Droplets are still entrained in the vapor down-
Fig. 10 provides a summary of flow regimes and heat transfer stream from the dry-out point, providing post-dry-out mist flow.
regimes for microchannels, which is based on findings from Eventually, all the droplets are consumed by evaporation and
both the saturated and subcooled boiling experiments and the the flow is converted into pure vapor. Notice in Fig. 10 how
above predictions. The existence of all the depicted regimes the heat transfer coefficient is both constant and fairly low in

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
LEE AND MUDAWAR: LOW-TEMPERATURE TWO-PHASE MICROCHANNEL COOLING 463

Fig. 11. Overall performance trends of small-diameter (50 m < d < 1000 m) two-phase microchannel heat sinks.

the inlet liquid region. The heat transfer coefficient begins in- not be possible because of low CHF. Overall, the highest
creasing at the point of incipient boiling and reaches a maximum possible cooling heat fluxes (i.e., highest CHF) are achieved by
around . This is followed by a reduction in in the sat- increasing both mass velocity and subcooling.
urated boiling region, reaching a minimum at the dry-out point. Fig. 11 also sheds some light on one of the most illusive
Fig. 10 also shows the variation of wall temperature along the aspects of fluid flow and heat transfer in small diameter chan-
microchannel. Lowest wall temperatures are encountered in the nels: the differences between microchannel and macrochannel
highly subcooled region and around . behavior. Generally speaking, conditions that yield very high
Notice that, while the indirect-refrigeration-cooling system void fractions for a given diameter, namely low mass velocity
can attain all of the flow regimes and heat transfer regimes de- and low subcooling, are more representative of “microchannel”
picted in Fig. 10 (assuming the microchannel is long enough), behavior. Conversely, conditions that yield relatively low void
the direct-refrigeration-cooling system can only sustain flow fractions for a given diameter, namely, high mass velocity and
and heat transfer regimes corresponding to . However, high subcooling, resemble those of “macro-channel” flow. The
the effectiveness of a microchannel cooling system depends not former “microchannel” extreme is were conventional two-phase
only on , but the ability to avoid CHF. Fig. 7 shows higher CHF regime maps cease to apply, and where interaction between
values are achieved with subcooled boiling. Subcooled boiling channels induces appreciable instability and premature CHF.
can generally be sustained by increasing inlet subcooling, in- The latter “macrochannel” extreme is where two-phase regimes
creasing mass velocity, and/or decreasing the microchannel’s are applicable, instabilities are inhibited, CHF is very high, and
length-to-diameter ratio. heat sink flow behavior follows that of single macrochannels.
From a system’s point of view, the indirect refrigeration Notice how increasing pressure decreases the void fraction,
system offers the advantages of greater flexibility in controlling pushing the entire system further towards the “macrochannel”
quality. In contrast, the compressor in a direct refrigeration extreme, while increasing surface tension produces larger
loop requires maintaining a microchannel exit quality of 1.0 bubbles and higher void fractions, pushing the system towards
or greater. Another practical attribute of the indirect cooling the “microchannel” extreme.
system is the ability to operate the microchannel heat sink at a
relatively mild pressure, compared to relatively high pressure V. CONCLUSION
for the direct cooling system, which is undesirable for elec- This study explored the use of low temperature refrigera-
tronic cooling because of structural concerns and added weight tion to maintain low device temperatures while dissipating high
of the cooling module. heat fluxes. Both direct- and indirect-refrigeration-cooling con-
Fig. 11 shows, in a mass velocity—subcooling plane, the figurations were examined. In the direct cooling system, a mi-
complex performance trends of a microchannel cooling system. crochannel heat sink serves as an evaporator in a conventional
Increasing mass velocity increases both the convective heat vapor compression cycle using R134a as working fluid. In the
transfer coefficient and CHF. However, these advantages are indirect cooling system, HFE 7100 is used to cool the heat sink
realized at the expense of greater pressure drop. The convective in a liquid loop that rejects the heat to a secondary refrigeration
heat transfer coefficient can also be increased by decreasing loop. Key findings from the study are as follows.
the inlet subcooling. Unfortunately, this also increases pressure 1) Two drastically different flow behaviors were observed
drop and, more importantly, operation at low subcooling may in these systems. Because of compressor performance

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
464 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 32, NO. 2, JUNE 2009

constraints, mostly high void fraction two-phase flow [6] V. Chiriac and F. Chiriac, “The optimization of a refrigerant vapor
patterns prevail in the R134a system, dominated by sat- compression system for power microelectronics,” in Proc. ITherm’06,
San Diego, CA, May-Jun. 30–6, 2006, pp. 759–764.
urated boiling. On the other hand, the indirect-refrigera- [7] R. Wadell, Y. K. Joshi, and A. G. Fedorov, “Experimental investiga-
tion-cooling system facilitates highly subcooled boiling tion of comapct evaporators for ultralow temperature refrigeration of
inside the microchannel heat sink. microprocessors,” ASME J. Electron. Packag., vol. 129, pp. 291–299,
2007.
2) Different pressure drop trends were observed between [8] W. Qu and I. Mudawar, “Measurement and prediction of pressure drop
the two systems. With R134a saturated boiling, pressure in two-phase micro-channel heat sinks,” Int. J. Heat Mass Transfer, vol.
drop increases with increasing heat flux, but this in- 46, pp. 2737–2753, 2003.
[9] M. B. Bowers and I. Mudawar, “High flux boiling in low flow rate, low
crease becomes milder as most of the flow is converted pressure drop mini-channel and micro-channel heat sinks,” Int. J. Heat
to vapor. Pressure drop with HFE 7100 subcooled Mass Transfer, vol. 37, pp. 321–332, 1994.
boiling first decreases with increasing heat flux because [10] M. B. Bowers and I. Mudawar, “Two-phase electronic cooling using
mini-channel and micro-channel heat sinks: Part 1- Design criteria and
of decreasing viscosity in the single-phase liquid re- heat diffusion constraints,” ASME J. Electron. Packag., vol. 116, pp.
gion. Pressure drop begins increasing following the 290–297, 1994.
commencement of boiling. Increasing mass velocity at [11] M. B. Bowers and I. Mudawar, “Two-phase electronic cooling using
mini-channel and micro-channel heat sinks: Part 2- Flow rate and
high fluxes actually decreases pressure drop because of pressure drop constraints,” ASME J. Electron. Packag., vol. 116, pp.
a reduction in void fraction. 298–305, 1994.
3) The convective heat transfer coefficient for the R134a [12] K. Mishima and T. Hibiki, “Some characteristics of air-water two-
and HFE 7100 systems follow opposite trends relative to phase flow in small diameter vertical tubes,” Int. J. Multiphase Flow,
vol. 22, pp. 703–712, 1996.
thermodynamic equilibrium quality. For R134a, the heat [13] K. A. Triplett, S. M. Ghiaasiaan, S. I. Abdel-Khalik, A. LeMouel, and
transfer coefficient is highest near and decreases B. N. McCord, “Gas-liquid two-phase flow in microchannels Part II:
monotonically with increasing . On the other hand, Void fraction and pressure drop,” Int. J. Multiphase Flow, vol. 25, pp.
395–410, 1999.
the convective heat transfer coefficient for HFE 7100 [14] T. N. Tran, M.-C. Chyu, M. W. Wambsganss, and D. M. France, “Two-
increases with increasing below . Highest phase pressure drop of refrigerants during flow boiling in small chan-
values are about equal for the two fluids. nels: An experimental investigation and correlation development,” Int.
J. Multiphase Flow, vol. 26, pp. 1739–1754, 2000.
4) While the R134a system can produce fairly large [15] H. J. Lee and S. Y. Lee, “Pressure drop correlations for two-phase flow
values, its cooling performance is limited by low CHF. within horizontal rectangular channels with small heights,” Int. J. Mul-
Because of its high CHF, the indirect cooling system tiphase Flow, vol. 27, pp. 783–796, 2001.
[16] M. Zhang and R. L. Webb, “Correlation of two-phase friction for re-
is far better suited for high-flux heat dissipation. Tests frigerants in small-diameter tubes,” Exp. Thermal Fluid Sci., vol. 25,
with this system yielded cooling heat fluxes as high as pp. 131–139, 2001.
840 W cm without encountering CHF. [17] A. Kawahara, P. M.-Y. Chung, and M. Kawaji, “Investigation of two-
phase flow pattern, void fraction and pressure drop in a microchannel,”
5) The results from both systems provide a global under- Int. J. Multiphase Flow, vol. 28, pp. 1411–1435, 2002.
standing of the cooling behavior of microchannel heat [18] W. Qu and I. Mudawar, “Flow boiling heat transfer in two-phase
sinks. These results are combined to construct a map micro-channel heat sinks – I. Experimental investigation and assess-
ment of correlation methods,” Int. J. Heat Mass Transfer, vol. 46, pp.
of performance trends relative to mass velocity, sub- 2755–2771, 2003.
cooling, pressure, and surface tension. Extreme condi- [19] W. Qu and I. Mudawar, “Flow boiling heat transfer in two-phase micro-
tions of near-saturated flow, low mass velocity, low pres- channel heat sinks – II. Annular two-phase flow model,” Int. J. Heat
Mass Transfer, vol. 46, pp. 2773–2784, 2003.
sure and high surface tension point to “microchannel” [20] J. Lee and I. Mudawar, “Two-phase flow in high-heat-flux
behavior, where macrochannel flow pattern maps fail micro-channel heat sink for refrigeration cooling applications:
to apply, instabilities are prominent, and CHF is quite Part I – Pressure drop characteristics,” Int. J. Heat Mass Transfer, vol.
low. On the other hand, systems with high mass velocity, 48, pp. 928–940, 2005.
[21] J. Lee and I. Mudawar, “Two-phase flow in high-heat-flux
high subcooling, high pressure, and low surface tension micro-channel heat sink for refrigeration cooling applications:
are far more stable and yield very high CHF values; Part II – Heat transfer characteristics,” Int. J. Heat Mass Transfer, vol.
two-phase flow in these systems follows the fluid flow 48, pp. 928–940, 2005.
[22] I. Mudawar, M. A. El-Masri, C. S. Wu, and J. R. Ausman-Mudawar,
and heat transfer behavior, as well as the flow pattern “Boiling heat transfer and critical heat flux in high-speed rotating liquid
maps of “macrochannels.” films,” Int. J. Heat Mass Transfer, vol. 28, pp. 795–806, 1985.
[23] I. Mudawar and M. B. Bowers, “Ultra-high critical heat flux (CHF)
for subcooled water flow boiling – I. CHF data and parametric ef-
REFERENCES fects for small diameter tubes,” Int. J. Heat Mass Transfer, vol. 42, pp.
[1] I. Mudawar, “Assessment of high-heat-flux thermal management 1405–1428, 1999.
schemes,” IEEE Trans. Compon. Packag. Technol., vol. 24, no. 2, pp. [24] J. Lee and I. Mudawar, “Fluid flow and heat transfer characteristics
122–141, Jun. 2001. of low temperature two-phase micro-channel heat sinks – Part 1: Ex-
[2] R. K. Kirschman, “Cold electronics: An overview,” Cryogenics, vol. perimental methods and flow visualization results,” Int. J. Heat Mass
25, pp. 115–122, 1985. Transfer, vol. 51, pp. 4315–4326, 2008.
[3] P. E. Phelan, V. A. Chiriac, and T.-Y. T. Lee, “Current and future [25] J. Lee and I. Mudawar, “Fluid flow and heat transfer characteristics
miniature refrigeration cooling technologies for high power microelec- of low temperature two-phase micro-channel heat sinks – Part 2:
tronics,” IEEE Trans. Compon. Packag. Technol., vol. 25, no. 3, pp. Subcooled boiling pressure drop and heat transfer,” Int. J. Heat Mass
356–365, Sep. 2002. Transfer, vol. 51, pp. 4327–4341, 2008.
[4] J. W. Peeples, W. Little, R. Shcmidt, and M. Nisenoff, “Low temper- [26] F. P. Incropera and D. P. Dewitt, Fundamentals of Heat and Mass
ature electronics workshop,” in Proc. 16th Semicond. Thermal Meas. Transfer, 5th ed. New York: Wiley, 2002.
Manag. Symp., San Jose, CA, 2000, pp. 108–109. [27] J. Lee and I. Mudawar, “Implementation of microchannel evaporator
[5] R. R. Schmidt and B. D. Notohardjono, “High-end server low-temper- for high-heat-flux refrigeration cooling applications,” ASME J. Elec-
ature cooling,” IBM J. Res. Develop., vol. 46, pp. 739–751, 2002. tron. Packag., vol. 128, pp. 30–37, 2006.

Authorized licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.
LEE AND MUDAWAR: LOW-TEMPERATURE TWO-PHASE MICROCHANNEL COOLING 465

Jaeseon Lee received the M.S. and Ph.D. degrees, Issam Mudawar received the M.S. and Ph.D.
both in mechanical engineering, from Purdue Uni- degrees from the Massachusetts Institute of Tech-
versity, West Lafayette, Indiana, in 2004 and 2008, nology, Cambridge, in 1980 and 1984, respectively.
respectively. His graduate research was focused on His graduate work involved magnetohydrodynamic
thermal management of high heat flux electronics (MHD) energy conversion and phase-change water
and fundamentals physics of two-phase flow and cooling of turbine blades.
heat transfer. He joined the Purdue University School of Me-
Currently, he is working at the United Technolo- chanical Engineering in 1984, where he established,
gies Research Center (UTRC), E. Hartford, CT, as a and became Director of, the Boiling and Two-Phase
Senior Research Engineer/Scientist. Flow Laboratory (BTPFL) and Purdue University
International Electronic Alliance (PUIECA). His
work has been focused on phase change processes, thermal management of
electronic and aerospace devices, intelligent materials processing, hydrogen
storage, high-Mach turbine engines, and nuclear reactor safety. His theoretical
and experimental research encompasses sensible and evaporative heating of
thin films, pool boiling, flow boiling, jet-impingement cooling, spray cooling,
microchannel heat sinks, heat transfer enhancement, heat transfer in rotating
systems, critical heat flux, and capillary pumped flows. He is also President of
Mudawar Thermal Systems, Inc., a firm that is dedicated to the development of
thermal management solutions.
Prof. Mudawar received several awards for his research accomplishments, in-
cluding the Best Paper Award in Electronic Cooling at the 1988 National Heat
Transfer Conference, the Best Paper Award in Thermal Management at the 1992
ASME/JSME Joint Conference on Electronic Packaging, the Journal of Elec-
tronic Packaging Outstanding Paper of the Year Award for 1995, and the Best
Paper Award in Thermal Management at ITherm 2008. He also received sev-
eral awards for excellence in teaching and service to Purdue students and their
organizations, including the Solberg Award for Best Teacher in School of Me-
chanical Engineering (1987, 1992, 1996, and 2004), the Charles Murphy Award
for Best Teacher at Purdue University (1997), and the National Society of Black
Engineers Professor of the Year Award (1985 and 1987). He was named Fellow
of the American Society of Mechanical Engineers (ASME) in 1998.

View publicationAuthorized
stats licensed use limited to: Purdue University. Downloaded on January 5, 2010 at 10:02 from IEEE Xplore. Restrictions apply.