Single- and Multiphase Flow for Electronic

Cooling

Yogendra Joshi and Zhimin Wan

Abstract
Liquid cooling is gaining critical importance for the continuing development of
high-density and high-power microelectronic system. In this chapter, both single-
phase and two-phase cooling are reviewed with a focus on various passive and
active heat transfer enhancement techniques including microchannels, wavy
channels, protrusions, dimples, bifurcating manifold, pin fin arrays, carbon nano-
tube, surface coatings, fluid mixture, and jet array. Although liquid cooling
demonstrates high heat transfer performance, its integration to the system pre-
sents great challenges in terms of cost and reliability issues. The difference of
system design between single-phase and two-phase cooling is described. Liquid
cooling is especially important for 3D stacked ICs. The codesign of architecture
and liquid cooling by considering the hot spot location, leakage power, and
routing is required for a successful implementation of liquid-cooled 3D IC
system.

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1 Why Liquid Cooling? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Indirect Cooling and Direct Immersion Cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Single-Phase Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Microchannel Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Micropin-Enhanced Microgaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3 System Configuration and Design Consideration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3 Thermal/Electrical Codesign of Microfluidic Cooled 3D ICs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.1 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2 Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

Y. Joshi (*) • Z. Wan

G.W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA,
USA
e-mail: [email protected]; [email protected]

# Springer International Publishing AG 2017 1

F.A. Kulacki (ed.), Handbook of Thermal Science and Engineering,
DOI 10.1007/978-3-319-32003-8_49-1
2 Y. Joshi and Z. Wan

4 Flow Boiling in Microgaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4.1 Characteristics of Pure Fluid Flow Boiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.2 Flow Boiling Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.3 Flow Boiling of Fluid Mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4 System Configuration and Design Consideration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6 Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

1 Introduction

The pursuit of multi-functionality and higher performance for computing, consumer,

and military microsystems, such as servers, tablets, radio-frequency
(RF) transceivers, solid-state lasers, and light emitting diodes (LED), continues to
push up the power consumption and surface area and volumetric power densities at the
chip and system levels. For example, gallium-nitride (GaN) high-electron-mobility
transistors (HEMT) have power densities many times higher than other power tran-
sistor technologies. The output power density of an N-polar GaN HEMTs with a
drawn gate length of 0.7 μm and a gate drain spacing of 0.8 μm can be 12.1W/mm
(Kolluri et al. 2011), resulting in a heat flux of 1.5
106 W/cm2. A pulse of sub-
picosecond duration and petawatt power was first reached at the Lawrence Liver-
more National Laboratory in 1996, and the Jan USP laser developed there has a peak
power intensity of 2
1020 W/cm2 (Salamin et al. 2006). Another common example
of a high heat flux device is the microprocessor for which “Moore’s law” (Moore
1998) has been the driving force for development (Mack 2011). The feature size has
become smaller and transistor count increased with each technology node to improve
integrated circuit (IC) functionality and performance, while decreasing costs (Intel
2016). A recent development in microelectronics is 3D stacked ICs. Compared to
conventional 2D ICs, the chips are stacked vertically, which greatly reduces the
interconnection length, a promising feature for the continuation of Moore’s law.
However, heat removal becomes a key challenge, due to the increased volumetric
heat generation and shrinking surface area.

1.1 Why Liquid Cooling?

For decades, air cooling has been used for cooling of various microsystems like
desktop computers, mobile phones, and LED systems. For mobile devices such as
smartphones, tablets, and smart watches, due to their compact size, natural convec-
tion air cooling is the most common cooling method. In higher-power computing
systems such as desktops, a copper or aluminum heat sink and fan are commonly
employed. Figure 1 shows a typical package cross section with backside air cooling.
Due to the low thermal conductivity of the underfill, substrate, and PCB, most of the
heat from the die (chip) is conducted through TIM1, lid, TIM2, and heat sink and
finally rejected to ambient air with the help of the fan. The junction-to-ambient
Single- and Multiphase Flow for Electronic Cooling 3

Fig. 1 Typical package stack with backside air cooling

thermal resistance includes the die, TIM1, lid, TIM2, heat sink base, and convection.
Key advantages of this cooling method are its low cost and high reliability.
Although air cooling is a cost-effective and reliable thermal management method,
the maximum allowable heat dissipation is limited. Webb (2005) suggested a
practical limit of thermal resistance value 0.336 K/W for a 16-mm2 heat source,
which gives 104 W heat rejection for impinging flow. Recent developments in
applications such as Intel server and high-performance computing have already led
to much higher heat dissipation (Intel 2016) beyond the capability of air cooling. If
not effectively removed, the high heat flux would result in high chip temperatures,
which will increase the parasitic or leakage power exponentially, degrade the
computational performance, and possibly even accelerate failure of the device
(Bakir and Meindl 2008). Liquid cooling offers heat transfer coefficients that are
one to two orders larger than in air cooling, allowing for continued development of
high-performance microsystem.
Thermal management is also extremely important for 3D IC development. As the
number of chips in a 3D stack increases, the package heat flux based on the top
surface area increases. The thermal resistances of the interior chips also increase due
to the stacking. Tier-based liquid cooling offers the promise of stacking an arbitrarily
large number of chips.

1.2 Indirect Cooling and Direct Immersion Cooling

For higher-power applications, the air-cooled heat sink can be replaced by a

coldplate with internal liquid flow, as shown in Fig. 2. The liquid circulates through
the coldplate, removing the heat from the package. The warm liquid is then pumped
via a remote heat sink, where it rejects the heat to the ambient, and returns cooled
liquid to the chip package. A variety of liquid coolants, including water, dielectric
4 Y. Joshi and Z. Wan

Fig. 2 Typical package stack with coldplate cooling

fluids, or refrigerants, can be used. Since the typical heat transfer coefficient for
liquid flow is much higher than for airflow for similar velocities and flow passage
dimensions, the size of coldplate can be reduced compared to air-cooled heat sink.
Although a fan is not required for the coldplate, a flow loop including pump,
reservoir, and liquid-air heat sink is needed. The liquid-air heat sink may be natural
convection cooled for lower power levels and fan cooled for higher powers.
Zhang et al. (2003) reported the thermal performance of a flip chip ball grid array
packages (FBGAs) with high heat dissipation (Fig. 3). Single-phase coldplate was
used for cooling with deionized water as the coolant. To further reduce the overall
thermal resistance, the package lid was removed, and the coldplate was placed
directly on the chip with a thermal interface material. The measured total thermal
resistance ranged from 0.44 K/W to 0.32 K/W for the 12-mm chip case and from
0.59 to 0.44 K/W for the 10-mm chip case, both under flow rates ranging from
1.67
106 m3/s to 1.67
105 m3/s.
A detailed layer-by-layer thermal resistance analysis revealed that the convection
resistance Rba for both cases was decreased by 55%, as the flow rate increased, while
the spreading resistance Rsp was decreased by 13% for the 12-mm chip case and 21%
for the 10-mm case. This decreased the total thermal resistance by 22% and 19%,
respectively.
One interesting finding is that the convection thermal resistance Rba decreases
quickly at low flow rate. But at higher flow rate, the decrease of Rba became slow and
saturated eventually with the further increase in flow rate. However, the pressure
drop increased quadratically with the increase of flow rate. On the other hand, the
chip resistance Rchip and TIM resistance RTIM are independent of flow rates. With the
decrease in coldplate heat sink resistance, RTIM became more significant, amounting
to as much as 50% of the total thermal resistance. So instead of continuing to
increase the flow rate, further reduction in the thermal resistances can be achieved
through the improvement of the TIM performance. Considerable work has been
done to improve the TIM performance by thinning TIM thickness and using high
Single- and Multiphase Flow for Electronic Cooling 5

Fig. 3 Cross-sectional view of the assembly of the coldplate on the flip chip BGA and the
corresponding thermal resistance

conductivity TIM materials with conductive particles in addition (Prasher 2006).

While the bulk thermal resistance of the TIM can be reduced by improving the
material properties, the contact resistance can still be significant, and degradation of
TIM material occurs due to aging. So the TIM performance has been a bottleneck of
the cooling of high-power chips. To reduce the TIM overall thermal resistance,
on-chip microfluidic cooling can be directly integrated with the chip (Dang et al.
2010), as seen in Fig. 4. The lid and TIM are completely removed from the
architecture in this approach. Fluid flows across the electrically inactive side of the
chip for heat removal. Usually a microchannel array, or a microgap with surface
enhancement structures, is etched on the back of the chips as the flow passage.
The advantages of on-chip microfluidic cooling are (1) the overall conduction
resistance is minimum; (2) the size of the cooling device is much smaller than the
coldplate and air-cooled heat sink solutions; and (3) heat transfer coefficient is
increased due to the reduced passage size, also requiring a smaller liquid inventory.
However, reliability concerns for on-chip microfluidic cooling result from the
pressure drop associated with liquid flow. For coldplate, the liquid does not contact
the chip directly, and structurally robust materials, such as copper, can be used to
fabricate the coldplate. For on-chip microfluidic cooling, high pressure drop can
pose a danger to the semiconductor chip. Fluid leakage risk is also a concern.
Overall, although microfluidic cooling has superior thermal performance, there are
numerous challenges associated with this new technology. The move from air
cooling to microfluidic cooling also requires a significant change in the design and
manufacturing processes. A key challenge is to integrate the microfluidic cooling at
low cost, without adversely impacting reliability. A closed flow loop with heat
exchanger, pump, and reservoir is needed for microfluidic cooling. These compo-
nents need to be packaged compactly, or the benefits of microfluidic cooling cannot
6 Y. Joshi and Z. Wan

Fig. 4 On-chip microfluidic cooling

Table 1 Comparison between three cooling solutions

Heat transfer System
Cost performance Reliability Size complexity
Backside air Low Low High Large Low
cooling
Coldplate cooling Medium Medium Medium Medium Medium
On-chip High High Low Low High
microfluidic cooling

be fully realized. Possible fluid leakage and lack of mechanical strength can cause
failure of the whole system and need to be addressed in a successful design. Table 1
shows a comparison between the three cooling solutions.
Liquid cooling is especially useful for future high-power electronics, especially
3D stacked ICs. However, before it can be employed, its characteristics need to be
fully understood. Significant work has been done during the past three decades to
characterize the single-phase and two-phase thermal performance of single micro-
channel and microchannel arrays. This work has been reviewed by Kandlikar
(2005), Steinke and Kandlikar (2006), Agostini et al. (2007), Kandlikar and Bapat
(2007), Khan and Fartaj (2010), Ebadian and Lin (2011), Mudawar (2011),
Kandlikar (2012, 2015), Adham et al. (2013), and Wu and Sunden (2014). With
the increasing interest in thermal management of emerging chip architectures, such
as three-dimensional (3D) stacked ICs, and interposer-based hybrid processor and
memory modules, microfluidic cooling is becoming of significant interest. Research
on cooling geometries, such as microgaps with surface enhancement structure, such
as micropin fin array, is relatively new, and the related work is limited. In this
chapter, the most recent developments in single-phase and two-phase microfluidic
cooling using microchannel and micropin fin-enhanced microgaps during the past
decade are reviewed in sections “Single Phase Convection,” “Thermal/Electrical
Co-design of Microfluidic Cooled 3D ICs,” and “Flow Boiling in Microgaps.” Focus
will be on the various thermal enhancement techniques. The system requirements
and challenges for both single-phase and two-phase cooling systems are identified.
The inter-tier liquid cooling is especially useful and compatible for 3D stacked ICs.
So the current development of 3D ICs with inter-tier microfluidic cooling and
electrical/thermal codesign of architecture and microfluidic cooling is reviewed.
Single- and Multiphase Flow for Electronic Cooling 7

Section “Summary” concludes the chapter with a summary of the existing work and
identification of gaps and future needs in microfluidic cooling to accelerate its
adoption.

2 Single-Phase Convection

Single-phase liquid cooling of microsystems has received extensive attention due to

its high thermal performance. This section reviews the characterization of thermal
transport within microchannels and microgaps with micropin fin array enhance-
ments. It also introduces various techniques to further improve the thermal perfor-
mance. The challenges in implementation of single-phase liquid-cooling system are
also discussed.

2.1 Microchannel Arrays

The microchannel heat sink cooling concept was first introduced by Tuckerman and
Pease (1981). A rectangular microchannel heat sink in a 1
1-cm2 silicon wafer
was fabricated. A Pyrex top cover was attached using anodic bonding to create a
flow path. The microchannels had a width of 50 μm and a depth of 302 μm and were
separated by a 50-μm-thick walls. Using deionized water as cooling fluid, the
microchannel heat sink was capable of dissipating 790 W/cm2 with a maximum
substrate temperature of 71  C above the water inlet temperature and a pressure drop
of 214 kPa. Due to their inherent advantages, microchannel array heat sinks have
received considerable attention (Samalam 1989; Harms et al. 1999; Rahman 2000;
Ochende et al. 2007; Naphon and Khonseur 2008) since Tuckerman and Pease’s
pioneering study.
The thermal and hydraulic performance of microchannels can be studied by
experiments, analysis, and numerical modeling. Numerical modeling is also an
important tool for optimization of microchannels, as part of the evaluation of new
designs. Qu et al. (2006) investigated flow development and pressure drop both
experimentally and computationally for adiabatic single-phase water flow. The
single rectangular microchannel was 222 μm wide, 694 μm deep, and 12 cm long,
and Reynolds numbers ranged from 196 to 2,215. A microparticle image
velocimetry system was used to measure velocity field at Reynolds number of
196 and 1,895 at several locations along the channel. A 3D computational model
was constructed to provide a detailed description of liquid velocity in both the
developing and fully developed regions. At high Reynolds numbers, sharp entrance
effects produced pronounced vortices in the inlet region that had a profound influ-
ence on flow development downstream. The prediction of the computational model
showed very good agreement with the measured velocity field and pressure drop.
Therefore, the conventional Navier-Stokes equations can accurately predict liquid
flow in microchannels and is a powerful tool for the design and analysis of micro-
channel heat sinks intended for electronic cooling. Lee et al. (2005) also validated
8 Y. Joshi and Z. Wan

Fig. 5 Effects of channel

width on thermal resistance
when pumping power is
0.01 W and 0.001 W (Adapted
from Kou et al. (2008))

the continuum theory with microchannel widths ranging from 194 to 534 μm and
depth being five times the width in each case. The microchannel was made of copper
and contained ten microchannels in parallel and deionized water was used as the
coolant. The results showed an average difference of 5% between experiments and
modeling for the Reynolds number ranging from approximately 300 to 3,500.
The effects of the channel height, width, and fin width on the heat transfer
coefficient and pressure drop have been studied and well established for micro-
channels. So the focus in this section is on the various thermal enhancement
techniques. Kou et al. (2008) built a 3D numerical model of a half microchannel
to study the heat transfer characteristics for various channel heights and widths. The
flow was assumed to be steady, incompressible, and fully developed. Constant fluid
properties and uniform heat flux boundary condition were applied.
Figure 5 shows the effects of the channel width on the thermal resistance when the
pumping power is fixed at 0.01 W and 0.001 W. An optimum width of the channel
can be observed. The optimum width is different for different pumping powers. The
sensitivity to channel width is more prominent at low pumping power.
Figure 6 shows the effect of microchannel height on the thermal resistance at
pumping powers 0.001 W and 0.01 W. The total heat sink height is fixed in this
study. Increasing the height of the microchannel decreases the substrate thickness. It
was found that the thermal resistance monotonically decreases with the increase of
the channel height. By using simulated annealing method, the global minimum
thermal resistance for a given pumping power can be found.
Another optimization method is the surrogate model optimization (Husain and
Kim 2008), as seen in Fig. 7. The first step is to define the design function and design
variables. Usually, the thermal resistance is the objective function. Two design
variables, ratio of channel width to height and ratio of fin width to height, were
selected for the optimization. Within the design variables’ range, a total of 16 design
points chosen from four-level full factorial design were used to construct the
surrogates. By solving Navier-Stokes and heat conduction equations, the thermal
resistances at specified design points can be obtained and optimized using surrogate
Single- and Multiphase Flow for Electronic Cooling 9

Fig. 6 Effects of channel

height on thermal resistance
when pumping power is
0.01 W and 0.001 W (Adapted
from Kou et al. (2008))

Fig. 7 Surrogate model optimization procedure (Adapted from Husain and Kim 2008)

models. Three surrogate models, response surface approximation (RSA), Kriging

(KRG), and radial basis neural network (RBNN), can be applied to predict the
optimal design point.
The assumption of fully developed flow in microchannel might not be valid. As
noted by Lee et al. (2005), the entrance effects should not be neglected. Experimen-
tal data were compared with both fully developed and developing laminar flow
prediction and significant deviations were seen. The chip width limits the total
number of channels. The previous optimization work did not include these effects
in their optimization. Wang et al. (2011) optimized the thermal resistance of a
microchannel heat sink with bottom surface area 10
10 mm. Constant heat flux
of 100 W/cm2 was applied in their model under developing flow conditions. Thermal
resistance was defined as a function of number of channels, N; channel aspect ratio,
α; and the ratio of channel width to pitch, β. The multi-parameter optimization
approach evaluated the gradients of the thermal resistance function and established
a conjugate direction for the updated search variables in each iteration. Under a
constant pumping power of 0.05 W, the optimal design values are N = 71, α = 8.24,
and β = 0.6, with a minimum overall thermal resistance of 0.144 K/W. One
important finding was that the optimized geometry depends on the pumping
10 Y. Joshi and Z. Wan

Fig. 8 Optimum geometric

parameters for various
pumping powers (Adapted
from Wang et al. (2011))

Fig. 9 Double-layer microchannel heat sink

power. As can be seen in Fig. 8, as pumping power increases from 0.005 to 1 W, the
optimal number of channels, N, and optimal channel aspect ratio, α, increases, and
the optimal ratio of the channel width to pitch, β, decreases.
Increase in heat removal ability of the microchannel heat sink can be achieved
either by increasing the heat transfer area or the heat transfer coefficient. To achieve
higher heat transfer area, multiple layers of microchannels can be stacked vertically
(Xie et al. 2013b). Figure 9 shows a double-layer microchannel. The flow direction
in each layer can be the same or in the opposite direction.
A three-dimensional model was built to study the heat transfer and pressure drop
characteristics of such double-layer microchannels under uniform heat flux dissipa-
tion. Compared to a single-layer microchannel, the heat transfer area of a double-
Single- and Multiphase Flow for Electronic Cooling 11

layer microchannel is doubled for the same single-layer microchannel size. So under
the same volumetric flow rate, the inlet velocity is half of a single-layer micro-
channel. Therefore, the pressure drop along the channel is reduced and thus the total
pumping power. The modeling results showed that to achieve the same thermal
resistance, the pressure drop of double-layer microchannel, either in parallel flow or
counter flow, is much smaller than the single-layer microchannel. This indicated that
double-layer microchannel can provide much better thermal performance than single
layer. Parallel-flow layout is better for heat dissipation at low pumping power, while
counterflow layout is better at high pumping power. This is because for counter flow
at low flow rate, the outlet fluid temperature in the first layer can be even higher than
the solid temperature at the second layer. So heat can be conducted from the liquid to
the solid and deteriorate the thermal performance. One advantage of the counterflow
layer is the reduced temperature gradient due to the mutual heat transfer between the
top and bottom channels. The channel height in each layer can be different. By
keeping the total height of the heat sink constant, while adjusting the height in each
layer, it was found that the thermal resistance also changes. This enables further
optimization of double-layer microchannel.
In a microchannel, the boundary layer develops as the fluid flows along it. Within
laminar developing flow regime, the heat transfer coefficient increases with Re. But
at the downstream region, the flow becomes fully developed, reaching a constant
value of the Nusselt number. Techniques to further enhance heat transfer include
increasing the flow mixing and promoting turbulence. Both passive and active flow
enhancement techniques have been investigated.
Passive heat transfer enhancement techniques include adding dimples, protru-
sions, and pins on the channel surface and using wavy channels, instead of straight
channels. Figure 10 shows rectangular microchannel with dimpled bottom surfaces.
Each microchannel is 50 μm deep and 200 μm wide (Wei et al. 2006). The in-line
dimples with a depth of 20 μm and footprint diameter of 98 μm are etched at the
bottom of the channel. A 3D CFD model was built to model one unit cell. Fully
developed periodic velocity and temperature boundary conditions were applied at
the inlet and outlet. Numerical modeling result revealed recirculating flow, and
secondary flow patterns, and their development along the flow direction. Heat
transfer augmentation (relative to a channel with smooth walls) is observed both
on the bottom dimpled surface and on the sidewalls of the channel. The pressure
drops in the laminar microscale flow were found to be either equivalent to or less
than that in a smooth channel with no dimples. Therefore the dimples, as an effective
passive heat transfer augmentation for macroscale channels, can be used for heat
transfer enhancement inside microchannels.
Flow characteristics and heat transfer performance in a rectangular microchannel
with combination of dimples and protrusions (Fig. 11) were also studied with water
(Lan et al. 2011). The results showed that incorporating dimples/protrusions in a
microchannel has the potential to enhance heat transfer, with low pressure drop
penalty. The normalized Nusselt numbers ranged from 1.12 to 4.77 and the
corresponding normalized friction factors from 0.94 to 2.03. The thermal perfor-
mance values showed that the dimple+protrusion cases perform better than the
12 Y. Joshi and Z. Wan

Fig. 10 Microchannel with dimples

Fig. 11 Microchannels with dimple/protrusion

smooth surface with dimples only. On the upstream portion of dimple, a lower heat
transfer region was produced because of flow separation. But on the downstream
portion of the dimple and in the dimple wake, high heat transfer was observed
because of flow reattachment and the vortices. Contrary to the dimples, a high heat
transfer region occurred on the upstream portion of the protrusion, while low heat
transfer region emerges on the downstream portion of the protrusion. In addition, the
cases with smaller streamwise pitches perform better. The staggered cases perform
better than the in-line ones.
Another passive heat transfer enhancement technique in a rectangular straight
microchannel heat sink is a bifurcation flow arrangement as shown in Fig. 12. The
bifurcating flow increases the heat transfer by breaking up the boundary layer
growth. Four different configurations with different bifurcation and length ratio
were studied (Xie et al. 2013a). Results showed that the channel with bifurcation
flow performed better than that of the corresponding straight microchannel.
The microchannel with larger bifurcation ratio and length ratio provided better
thermal performance with higher pressure drop. So by proper design of the bifurca-
tion length ratio and number of channels, the overall thermal performance of the
liquid-cooling microchannel heat sinks can be maximized.
Adding pin fins within a microchannel also enhances heat transfer by increasing
the surface area and enhances flow mixing. The performance of a microchannel with
offset pin fins inside was compared with a typical microchannel with no pins
(Shafeie et al. 2010). It was shown that using offset pin fins increases heat removal
rate. However, the pressure drop is also significantly increased at the same time,
which deteriorates the overall coefficient of performance for microchannel with this
Single- and Multiphase Flow for Electronic Cooling 13

Fig. 12 Straight
microchannel heat sink with
bifurcation flow arrangement

Fig. 13 Wavy microchannel

type of microstructure. The heat removal and pressure drop increase 22% and 83%,
respectively.
Wavy microchannel (Fig. 13) has also been studied due to the secondary flow and
vortices, leading to chaotic advection, which can greatly enhance the convective
fluid mixing and thus the heat transfer performance (Sui et al. 2010). A comparison
with straight microchannels for the same cross section showed that wavy micro-
channels provide significant improvement in performance. In addition, the pressure
drop penalty of the wavy microchannels is much smaller than the heat transfer
enhancement. Furthermore, the wavy microchannel design is flexible in terms of
relative wave amplitude and waviness. The wave amplitude of the microchannels
can be varied along the flow direction for various practical purposes, without
compromising their compactness and efficiency. By increasing the waviness along
the flow direction, higher heat transfer performance can be achieved, increasing the
temperature uniformity across the devices. Lastly, high waviness can be designed in
the high heat flux regions to mitigate hot spot.
One of the active heat transfer enhancement techniques is an impinging jet
(Robinson 2009). The coolant is injected onto the hot surface from an orifice. The
heat transfer performance of an impinging jet array was compared with micro-
channel (Robinson 2009). Both heat sinks were required to dissipate 250 W/cm2
while being maintained at a temperature of 85  C (Fig. 14). The modeling results
showed that both the impinging jet array and microchannel heat sinks can meet the
cooling requirements with pumping power less than 0.1 W. Microchannels achieve
this cooling target at high pressure drop and low volumetric flow rate. On the
14 Y. Joshi and Z. Wan

Fig. 14 Schematic of microchannel and impinging jet array

contrary, impinging jet array heat sink requires a lower pressure drop and higher
volumetric flow rate. For a cooling system, lower pressure drop is favorable with the
concern of system reliability. Higher flow rate can reduce the bulk fluid temperature
rise and thus increase the temperature uniformity across the device. Another benefit
of impinging jet array is that the orifices can be concentrated on the high heat flux
area, while microchannel is a global cooling strategy. From a practical engineering
point, the cost of impinging jet array heat sink can be lower since the orifice plates
can be manufactured at a reduced cost due to its simplicity. So impinging jet array
heat sink can be the choice of future high-power electronic devices.
The benefits of microchannels with high surface area and impinging jet arrays
with high heat transfer performance can be combined to produce a hybrid cooling
module in which a series of micro-jets deposit the coolant into each microchannel in
a heat sink (Sung and Mudawar 2008). The coolant absorbs the heat and is expelled
through both ends of the microchannel (Fig. 15). Three microjet patterns were
studied: decreasing jet size (relative to center of channel), equal jet size, and
increasing jet size. The heat transfer performance was evaluated with dielectric
fluid HFE 7100 as the coolant. 3D numerical simulation using the standard k-e
model predicted the experimentally measured wall temperatures well. The modeling
results showed that the hybrid cooling module presented complex interactions
between impinging jets and microchannel flow. When increasing the coolant flow
rate, the heat transfer performance was enhanced, and thus wall temperature is
decreased. However, this benefit is achieved at the penalty of greater wall temper-
ature gradients. The comparison of the three patterns showed that decreasing-jet-size
pattern yields the highest convective heat transfer coefficients and thus lowest wall
temperatures, while the equal-jet-size pattern provides lowest temperature gradient.
The increasing-jet-size pattern resulted in greater wall temperature gradients. This is
because the impingement from larger jets near the channel outlets blocks the spent
fluid flow.
In the previous hybrid cooling module, high-temperature gradient occurs as the
fluid flows to the two ends. A modified high-performance ultrathin manifold micro-
channel heat sink is shown in Fig. 16 (Escher et al. 2010). The heat sink consists of
Single- and Multiphase Flow for Electronic Cooling 15

Fig. 15 Hybrid cooling module. (a) Equal jet size. (b) Decreasing jet size. (c) Increasing jet size

impinging liquid slot jets on a structured surface fed with liquid coolant by an
overlying two-dimensional manifold. The cool liquid is injected into the slot and
returned immediately to the upper layer through nearby slots, which significantly
reduce the bulk fluid temperature rise. It was found that a design with 12.5 manifold
systems and 25-μm-wide microchannels as the heat transfer structure achieved
optimum performance. The chip size is 2 cm and 2 cm and water is the coolant.
The thermal test chip is 500 μm thick and the microchannel is 300 μm deep, while
the total height of the heat sink does not exceed 2 mm. With a volumetric flow rate of
1.3 l/min, a total thermal resistance between the maximum heater temperature and
fluid inlet temperature of 0.09 cm2K/W was achieved with a pressure drop of
0.22 bar. So maximum temperature difference between the chip and the fluid inlet
is 65 K when the power density is more than 700 W/cm2.
The high performance of the manifold structure allows for elevated inlet temper-
ature of up to 70  C. So for a flow rate of 1 l/min and heat flux 100 W/cm2, the
16 Y. Joshi and Z. Wan

Liquid
Liquid Liquid

Fig. 16 Schematic of a manifold microchannel heat sink

maximum chip temperature can be maintained at 81  C. The low thermal resistance

and the tolerable high inlet temperatures promise the possibility of eliminating the
need for chillers. In addition, the rise of the inlet temperature to 70  C enhances the
overall efficiency of the heat sink by more than 40%. Hence, not only the energy for
the chiller can be saved but also decrease the total pressure drop across the heat sink
and thus increase the overall performance of the cooling system. The performance of
the manifold heat sink can be further enhanced by lowering the hydrodynamic
resistance of the heat sink. This can be achieved by decreasing the nozzle length
and by increasing the manifold channel height. The reduction in the hydrodynamic
resistance of the manifold results in higher numbers of manifold systems, reducing
the flow path length through the heat transfer structure. Therefore, narrower channel
widths can be designed and further decrease the total thermal resistance of the
heat sink.
Another modified design of the microjet array is parallel inlet and outlet mani-
folds with a distributed return (Brunschwiler et al. 2006). The nozzle to heater gap
size ranges from 3 to 300 μm. Pressure drops less than 0.1 bar at 2.5 l/min flow rate
have been obtained with a hierarchical treelike double-branching manifold. An
optimal heat-removal rate of 420 W/cm2 using water as a coolant was achieved
through parametric study. For a near-optimal design with a gap to inlet diameter ratio
of 1.2, a heat transfer coefficient of 8.7 W/cm2 K and a junction to inlet fluid unit
thermal resistance of 0.17 Kcm2/W were obtained. This allows for 370 W/cm2
cooling performance at a junction to inlet fluid temperature rise of 63 C, a pressure
drop of 35 kPa, and a flow rate of 2.5 l/min.
High heat transfer performance in a microchannel can also be achieved by using
high conductivity substrate material. Due to the superior thermal conductivity of
individual carbon nanotube (CNT) in the axial direction, CNT arrays have been
explored as an effective material for the thermal management of MEMS-based
devices by introducing CNT into the microchannel (Ekstrand et al. 2005). A
reduction of thermal resistance of the microchannel from 0.98 K/W to about
0.43 K/W was obtained when CNT array fins were used. One concern about the
Single- and Multiphase Flow for Electronic Cooling 17

Fig. 17 Split-flow design

Fig. 18 Bifurcating tree-

shaped microchannel

CNT is its strength. By utilizing a metal-enhanced CNT transfer technique, the

interface between the CNTs and the chip surface was improved by minimizing the
thermal contact resistance and promoting the mechanical strength of the microfins. A
heat flux of 1,000 W/cm2 with water was projected due to the increased heat
exchange area of the CNT microfins (Fu et al. 2012).
The bulk fluid temperature rise in microchannel results in high chip temperature
nonuniformity. To reduce the bulk fluid temperature rise, a split-flow channel
(Fig. 17) was proposed (Tan et al. 2011). The advantages of the split-flow arrange-
ment over the single-pass arrangement are (1) for the same pumping power, the total
flow rate can be doubled, while the pressure drop halved due to the reduced flow
path, and (2) the bulk fluid temperature rise along the microchannel is also halved if a
uniform heat flux is assumed. By designing the heat sink with a supply plenum
tapering downstream, an even flow distribution for removing the heat away from the
die can be obtained to further improve the thermal performance. With such a
microchannel array arrangement, a thermal resistance of 0.15 K/W and lower
temperature variation on die can be obtained. The lower pressure drop due to the
shorter flow length allows integrated pumps to operate either at smaller sizes or
higher flow rates. Kandlikar and Upadhe’s experiment with the split microchannel
showed dissipating heat fluxes beyond 300 W/cm2 using water as the coolant with a
pressure drop of around 35 kPa (Kandlikar and Upadhye 2005).
Tree-shaped microchannel patterns (Fig. 18) have been pursued to improve the
chip temperature uniformity (Wang et al. 2007). The effect of the bifurcation angles
in the constructal nets on the fluid flow and heat transfer characteristics of such
network were studied using a three-dimensional computational fluid dynamics
18 Y. Joshi and Z. Wan

Fig. 19 Bifurcating manifold

approach. Results show that the bifurcation angle is an important factor for such
cooling nets. Surface temperature distributions and pressure drop along the flow
paths are analyzed and compared. For the same boundary conditions, a lower
temperature and pressure variation are observed at lower bifurcation angles.
The hydrodynamic and thermal performance of a bifurcating (Fig. 19) and a
parallel channel network branching from a single manifold channel were compared
(Escher et al. 2009). Although it has larger temperature nonuniformity for the
parallel microchannel, it has also several advantages: (1) the parallel channel is
much more densely packed than the bifurcating manifold and distributes the coolant
more efficiently to a larger heat transfer surface area; (2) for a constant flow rate, the
parallel channel network achieves more than 5
higher coefficient of performance
than the bifurcating manifold, while almost 4
more heat can be removed for a
constant pressure gradient; and (3) parallel channel design is less complicated and
cheaper compared to the bifurcating manifold. The parallel channel is integrated into
a single plane, while the bifurcating design needs a second plane for fluid return.

2.2 Micropin-Enhanced Microgaps

Significant temperature variations across the chip persist for conventional single-
pass parallel-flow microchannel heat sink, since the heat transfer performance
deteriorates in the flow direction in microchannels, as the coolant heats up. Recent
advancement in micromachining techniques allows more complex 3D microsize
geometries to be fabricated directly into the high thermal conductivity materials that
can be used as the substrates for miniature heat sinks. This makes it possible to
explore structures that may be more effective in heat transfer enhancement than the
parallel microchannels and can provide better temperature uniformity. One such
enhanced structure is micropin fin arrays (Fig. 20), with pin characteristic dimen-
sions of tens to hundreds of micrometers and height (Hp) to diameter (Dp) ratio from
0.5 to 8. The flow disruptions provided by the separated pin fins increase the flow
Single- and Multiphase Flow for Electronic Cooling 19

Fig. 20 Schematic of
staggered micropin fin array

mixing and can also serve to break up the boundary layer (Steinke and Kandlikar
2004).
Thermal design and performance assessment of a micropin fin heat sink require a
fundamental understanding and accurate prediction of flow and heat transfer in
microsize short pin fin arrays. The thermal-hydraulic performance of micropin fin
arrays and microchannel was compared in terms of total thermal resistance (Rtot) by
Peles et al. (2005). At an inlet pressure of two atmospheres, the minimum Rtot was
0.0389 K/W with water, which corresponded to 7.8  C maximum wall temperature
raise for 200 W/cm2 heat flux, and 30.7  C at 790 W/cm2, compared to 71  C for
microchannel cooling (Tuckerman and Pease 1981). A recent study by Jasperson
et al. (2010) showed that the flow rate was a factor in determining whether micro-
channels or micropin fin arrays have better performance. In their study, micro-
channels and micropin fin arrays of same height and width (670 μm and 200 μm)
were made of copper. Under a mass flow rate from 30 to 90 g/min, the pressure drop
of micropin fin heat sink was always higher than that of microchannel heat sink, the
difference increasing with the flow rate. The convection thermal resistance, Rconv, of
micropin fin heat sink was higher than that of microchannel heat sink at flow rate less
than 60 g/min and lower above 60 g/min.
The micropin fin arrays can be arranged in in-line or staggered configurations.
Kosar et al. (2005) experimentally studied and compared the heat transfer coefficient
associated with the forced flow of deionized water over staggered and in-line circular
micropin fins. The Dp was 100 μm, Hp was 100 μm, and longitudinal (SL) and
transverse (ST) spacings were both 150 μm. Under the same Re, the staggered
configuration resulted in higher heat transfer coefficient than the in-line configura-
tion. A more recent comparison study by Brunschwiler et al. (2009) showed that
under the same flow rate, in-line pin fin showed lower heat transfer coefficient and
higher thermal resistance than staggered pin fin. Another interesting finding was that
in-line pin fin presented a flow regime transition, which manifested itself as an abrupt
20 Y. Joshi and Z. Wan

pressure gradient change and a local heat transfer maximum. The transition moved
toward the inlet at increasing flow rate.
Due to the superior performance of staggered micropin fin arrays, considerable
research has been done to study its thermal and hydraulic characteristics, including
the effects of Re, ST, SL, Hp, pin shape, and tip clearance (tc).

2.2.1 Effect of Re
Due to the small pin fin dimensions at the microscale, the flow regime is expected to
be predominantly laminar. For a specific pin fin configuration, the heat transfer
coefficient (h) and friction factor increase with Re. The nondimensional friction
factor (f) could be defined as (Kosar and Peles 2007):

2ΔPρf
f fan ¼ (1)
N L G2
where ρf is the fluid density, NL is the number of pins in the longitudinal direction,
and G is the maximum mass flux at the smallest cross-sectional area.
The friction factor is comprised of two components, one accounting for the drag
due to flow separation and the other stemming from shear stress (Kosar and Peles
2007). As the fluid flows across the micropin fins, a thin boundary layer is formed at
the pin surface. As Re increases, the boundary layer becomes thinner, and flow
separation is enhanced (Kosar and Peles 2007). Kosar et al. (2005) experimentally
studied the pressure drop and f of circular micropin fin arrays. Their results showed
that f decreased with increasing Re, which was also observed by Prasher et al.
(2007), Short et al. (2002b), and Tullius et al. (2012). A change in the relationship
of f to Re was observed by both Kosar et al. (2005) and Prasher et al. (2007),
showing that f was very sensitive to Re for Re < 100 and was less sensitive to Re for
Re > 100. It was believed that a flow pattern transition occurred around Re = 100.
The Nusselt number (Nu) increased with Re, which was also observed by Prasher
et al. (2007), Tullius et al. (2012), and Short et al. (2002a). Similarly, two distinct
regions of the Nu dependency on the Re separated by Re = 100 have been identified
(Kosar et al. 2005; Kosar and Peles 2007). Kosar and Peles (2006a, b) attributed such
dependency to two factors: endwall effects and a delay in flow separation. Flow
separation was assumed to control the transition Re.

2.2.2 Effect of Spacing

The effect of longitudinal spacing was evident when comparing the friction factors
of micropin fin array with SL = 150 μm and SL = 350 μm (Prasher et al. 2007).
Device with SL = 150 μm had larger f than device with SL = 350 μm, which
suggested that densely populated pin fins lead to higher f. This was because that
wakes behind pin fins formed due to flow separation. For densely populated pin fins,
the wake generated downstream of a pin fin interacted more strongly with pin fins in
the following row. As a result, f were higher for closely packed objects. This was
also observed by Prasher et al. (2007) and Short et al. (2002b). However, Tullius
et al. (2012) showed an opposite trend. h of device SL = 150 um were greater than for
Single- and Multiphase Flow for Electronic Cooling 21

device SL = 350 μm over Re ranging from 30 to 112 (Kosar and Peles 2007). This
may be due to the pronounced wake-pin fin interaction in SL = 150 μm. Because of
the dense spacing, the wake formed downstream a pin fin may interact with the pin
fins in the following row, so that mixing and heat transfer were enhanced. This was
reflected as higher h in SL = 150 μm. Since at low Re, the wake-pin fin interaction is
moderated, deviations between Nu of the two devices diminished. Results of Prasher
et al. (2007) agreed with observation in (Kosar and Peles 2007). However, Both
Tullius et al. (2012) and Short et al. (2002a) showed that sparse pins had larger h.
Very little work studied the effect of ST on f and Nu. Prasher et al. (2007) and
Short et al. (2002b) reported an increasing f with decreasing ST, while an opposite
trend was presented by Tullius et al. (2012). Both Tullius et al. (2012) and Short et al.
(2002a) found that Nu increased with increasing ST.

2.2.3 Effect of Aspect Ratio (Hp/Dp)

The ratio of Hp to Dp has a significant effect on the f. It was reported that the
micropin fin arrays with lower HP/DP ratio produced higher f at the same pin
densities and Re (Kosar et al. 2005). However, The HP/DP ratio effect reduced
with increasing Re (Kosar et al. 2005). This was attributed to the endwall effect.
Small aspect ratio devices were more affected by the hydrodynamic boundary layer
imposed by the endwalls (wall pin interaction), resulting in increased viscous shear
forces, and therefore larger f were obtained (Kosar et al. 2011). The effect of pin fin
Hp to Dp aspect ratio was significant for low Re and diminished at larger Re because
of a thinning hydrodynamic boundary layer. Therefore, with the increase in Re, the
deviation between smaller and larger Hp to Dp pillars moderated. The observation in
(Kosar et al. 2005, 2011) agreed with Prasher et al. (2007) for Re <100 and Short
et al. (2002b). However, f was reported to be increased with aspect ratio for
Re > 100 (Prasher et al. 2007; Tullius et al. 2012). The dependence of the Nu on
the Re was considerably more notable than for long pin fin as a result of endwall
effects (Kosar et al. 2005) since the endwall effects were significant at a low Re flow
over short pin fins (1/2 < HP/DP < 6). Two intrinsically coupled physical factors
adversely affected h at 10 < Re < 100 for flow over micropin fins: the thermal and
hydrodynamic boundary layer established at the endwall and a delay in flow
separation to higher Re. The suppression of h was amplified with the reduction of
H/D ratio and was more evident at low Re (Kosar and Peles 2006a). The results of
Short et al. (2002a) showed that Nu decreased with increased aspect ratio. However,
Koz et al. (2011) did a parametric study on the effect of endwalls on heat transfer and
found that Nu increased with Re but not necessarily with HP/DP. Nu of various HP/
DP ratios can show different trends when plotted against Re. Tullius et al. (2012)
found that Nu increased with aspect ratio.

2.2.4 Effect of Tip Clearance

The clearance (tc) between the pin tip and wall was believed to affect the overall
hydraulic and thermal performance of a given cooling system (Sparrow et al. 1980).
Introducing tip clearance resulted in a decrease in form drag and an increase in shear
stresses at the walls (Rozati et al. 2008). The balance between the two determined the
22 Y. Joshi and Z. Wan

overall increase or decrease in f. A slight increase in f with tc/DP = 0.1 at high Re

was observed (Sparrow et al. 1980). As the tc increased further to tc/DP = 0.3 and
0.4, f was decreased. The reduction in f with increasing tc was much more evident at
higher Re. Moores et al. (2009) also found that the overall pressure drop is initially
increased with the introduction of tip clearances for tc/DP < 0.1 and then decreased.
The pin fins tend to have relatively low height to diameter ratio due to the high
h. So the pin tip area can represent a considerable portion of the total area of the
array. Exposing the tips of the pin fins to the cooling fluid can increase the total heat
transfer area. Also the tip clearance introduces a three-dimensional behavior into the
flow field around the pins and may change the local heat transfer rate. Moores and
Joshi (2003) reported an increase of heat transfer rate with the introduction of tc and
attributed the enhancement primarily to the additional surface area exposed to the
cooling fluid. Rozati et al. (2008) believed that tc affected h by eliminating viscosity-
dominated endwall effects on the pin, by eliminating the pin wake shadow on the
endwalls, by inducing accelerated flow in the vicinity of the top wall and the pin top,
by reducing or impeding the development of the recirculating wakes, and by
redistributing the flow along the height of the channel. For a tip gap of tc/DP = 0.1,
Nu decreased sharply. As tc/DP continued to increase, the Nu increased (Rozati et al.
2008).

2.2.5 Effect of Pin Shape

Advanced microfabrication technologies enable researchers to explore the charac-
teristics of micropin fins with different shapes besides circular pins. The pins can be
divided into two categories: streamlined shapes, such as hydrofoils, and
unstreamlined ones including circular, square, diamond, cone, triangle, hexagon,
and ellipse. Heat transfer characteristics of square micropin fins were experimentally
studied by Siu-Ho et al. (2007) and Liu et al. (2011). They found that both heat
transfer coefficient and pressure drop increased with increasing Re. Also, h was
higher near the heat sink inlet and decreased along the flow direction. This trend may
be caused by entrance and streamwise heat conduction effects (Siu-Ho et al. 2007).
A friction factor transition phenomenon appeared at Re ~ 300 (Liu et al. 2011).
Micropin fins having sharp edges generated higher h than streamlined pin fins
(Kosar and Peles 2007). This was associated with separation effects mitigated by
sharp edges, as well as extended wake regions, which increased the mixing and heat
transfer. However, sharp-edged pins also resulted in larger f, at the same Re. The
sharp edges enhance wake-pin interactions, causing larger pressure drops. A com-
parison between circular, diamond, and square micropin fins with same hydraulic
diameter, spacing, and height by Kosar et al. (2005) and Mita et al. (2011) showed
that for the same Re, circular micropin fins had smallest f. Hydrofoil pin fins with
zero angle of attack resulted in considerably lower f compared to circular pins,
especially at high Re (Kosar et al. 2011), around 7.5 times lower at a Re of 1,000.
The difference in f continued to amplify as the Re increased above 1,000, primarily
because crossflow over circular-shaped fins transitioned to turbulent flow, where the
friction factor is less dependent on Re, while flow over the hydrofoil fins was well
within the laminar flow regime. Therefore, the merits of using micropin fin device
Single- and Multiphase Flow for Electronic Cooling 23

are dependent on the performance evaluation criterion used, as well as on the

hydrodynamic conditions. In general, for a fixed pressure drop and pumping
power, utilizing streamlined pin fin heat sink is favored at moderate pressure drops
and flow rate, while for very high and lower pressure drops and flow rates, pin fins
promoting flow separation should be favored. For fixed mass flow rate, streamlined
pin fins provided inferior performance.

2.3 System Configuration and Design Consideration

Attention was focused on the performance of the microchannel itself in the previous
section. However, the successful integration of the microchannel heat sink in the
overall cooling system is equally important but has received limited attention.
Figure 21 shows a typical liquid-cooling system. It consists of pump, controlled
temperature bath, flow meter, filter, metering valve, pressure transducer, thermocou-
ple, microchannel heat sink, air-cooled heat sink, and reservoir. The pump circulates
the fluid through the microchannel heat sink. A filter is used to keep the inlet water
clean and prevent clogging of the microchannels. An air-cooled remote heat sink is
used to reject the heat to the ambient. The reservoir is used to store excess fluid
inventory to make up any deficit over extended operation periods.
Based on the thermal and pressure drop performance of the microchannel heat
sink, and the chip power dissipation, the necessary flow rate within the cooling
system can be determined. Next, the remote heat sink, pump, and reservoir can be
selected based on the flow rate. The component materials must be compatible with
the coolant, which can be either water, often with additives, or dielectric fluid,
depending on the applications and the dissipated heat. If the microchannel heat

Fig. 21 Typical liquid-cooling system

24 Y. Joshi and Z. Wan

sink is used as an externally attached coldplate, metals such as copper can be used to
fabricate it, and it can withstand higher pressures than a silicon microchannel heat
sink. Typically, water which has excellent convective heat transfer characteristics
can be used as the coolant. For on-chip microchannels, dielectric fluid may often be
required, to mitigate any danger of electrical shorts. Before the coolant is charged
into the flow loop, it needs to be degassed, and the flow loop vacuumed. Non-
condensable gas in the loop can cause malfunction of the pump and decrease the
flow rate. It can also degrade the convective heat transfer coefficient.
A chip such as a microprocessor may work continuously for several years.
Therefore, a highly reliable cooling system is required. First, the flow loop should
be leakage-free. Any leakage of the liquid will reduce the amount of the coolant in
the flow loop and eventually lead to failure of the processor. Secondly, the flow loop
should be very clean. A filter is usually placed upstream of the microchannel heat
sink. If the pore size of the filter is too small, the filter can be blocked by contaminant
particles in the loop, reducing the flow rate. Thirdly, the microchannel heat sink
should be designed to withstand the mechanical stresses resulting from the fluid
flow. High pressure can cause cracking of the microchannels in semiconductor
materials such as Si. So the microchannel heat sink should be carefully designed
to reduce the pressure drop. In addition, a compact cooling system is required.
Although the size of the microchannel heat sink is much smaller than conventional
air-cooled heat sink due to its high performance, the other components typically
increase the size and complexity of the whole system. Lastly, the cost of the liquid-
cooling system must be considered. A liquid-cooling system is usually more expen-
sive than an air-cooled heat sink. The performance benefit gained by the liquid-
cooling system must exceed the extra cost due to the liquid-cooling system.
Among many, the most prominent but often ignored challenge is the long-term
reliability of the system. In order to be practical, the thermal management system
must maintain its cooling efficiency for the entire duration of service time. Any
physical or chemical degradations of the coolant itself or the components of the
system can adversely impact the cooling efficiency and, therefore, must be prevented
(Chang et al. 2006). In this regard, corrosion is particularly problematic for liquid-
based thermal management for microsystems with an integrated microscale heat
exchange system. While water as a coolant provides superior heat transfer proper-
ties, it tends to corrode components of the thermal management system, particularly
in the presence of additives such as antifreeze. Additives must not seriously impair
the desirable thermal transport properties of water, such as low viscosity and high
heat capability, while protecting it from freezing and posing no serious health hazard.
Majority of antifreeze additives are based on ionic salts, most notably calcium
chloride, potassium acetate, and sodium formate. Their addition increases the chem-
ical activity of the coolant, thereby the probability of corrosion. Secondly, the ratio of
surface area to volume can reach as high as 1cm2/ml, which means that even a slight
amount of corrosion can seriously affect the basic properties of the coolant and result
in an acceleration of corrosion. Finally, corrosion products can significantly endan-
ger the system reliability by causing clogging and thus rapidly decreasing the
cooling efficiency.
Single- and Multiphase Flow for Electronic Cooling 25

It is very important that a complete assessment of the corrosion risk for a chosen
coolant is conducted before implementation of liquid-cooling system. Equally
important is to find suitable coolant additives that can effectively prohibit or retard
the corrosion.

3 Thermal/Electrical Codesign of Microfluidic Cooled 3D ICs

3D ICs can significantly increase the processor performance by vertical stacking of

chips. However, when the chips are stacked, the heat flux per unit surface area is
increased, and the interior layers in the stack are susceptible to overheating. On-chip
liquid cooling is a promising choice for thermal management of 3D ICs. The heat
from the chip is removed by individual liquid-cooling layers and therefore the
number of chips stacked is not limited. A lot of work has been done to study the
thermal performance of 3D ICs with liquid cooling, experimentally and numerically.

3.1 Experiments

Sekar et al. (2008) presented a 3D IC technology with integrated microchannel

cooling (Fig. 22). Fluidic interconnect network fabrication proceeds at the wafer
level, is compatible with CMOS processing and flip-chip assembly, and requires four
lithography steps. The electrical signal vias are embedded in the fins of microgaps.
Experimental results show that the integrated microchannel heat sinks enable
cooling of power density of more than 100 W/cm2 of each chip tier. A key

Fig. 22 3D ICs with integrated microchannel

26 Y. Joshi and Z. Wan

Fig. 23 Integrated cooling solution for a compact 3D silicon module

characteristic of this platform is the ability to assemble chips with electrical and
fluidic I/Os and seal fluidic interconnections at each stratum interface (King et al.
2010).
Tang et al. (2010) proposed an integrated cooling solution for a compact 3D
silicon module (Fig. 23), which spreads the heat over a larger coldplate heat
exchanger, where conventional air-cooling suffices to reject the heat. The system
is designed such that no external fluidic interconnects are required to assemble the
package on the mother board. Two silicon carriers with embedded fluidic micro-
channels are stacked vertically, and each carrier is mounted with a chip dissipating
100 W. At a high flow rate (400 ml/min) for a single carrier, the maximum pressure
drop in the carrier is 90 kPa. The sealing technique for the fluidic path worked well
and no leakage was observed. The pressure drop can be further reduced by the
optimized design and a decrease of 30–50% was obtained, as compared to a same-
sized convectional carrier. The simulation results showed that the thermal resistance
of the optimized carrier and the conventional carrier are each about 0.18 K/W.
However, the variation of the junction temperature for the optimized carrier is
8.1 K, which is much lower than 14.1 K for the conventional carrier.

3.2 Modeling

Numerical modeling enables rapid evaluation of thermal performance, and therefore

a significant work has been done to characterize the performance of 3D ICs with
integrated liquid cooling. Koo et al. (2005) presented a 3D circuit network to study
the effect of logic and memory layout on the thermal performance. Their results
Single- and Multiphase Flow for Electronic Cooling 27

show that integrated microchannel cooling can be scaled up to multiple tiers.

Aravind presented 3D ICE (Sridhar et al. 2010), a compact transient model for the
thermal simulation of 3D ICs with multiple inter-tier microchannel liquid cooling.
The model is compatible with existing thermal CAD tools for ICs and demonstrated
significant speedup (up to 975
) over typical commercial computational fluid
dynamics (CFD) simulation tool, while preserving reasonable accuracy. Alfieri
et al. (2010) developed a multiscale heat transfer model for integrated water cooling
of a four-tier chip stack with a footprint of 1 cm2. The cooling structure, which
consists of microgaps with cylindrical pin fin array, can be directly integrated with
the device layout in multilayer chips. The total power in the stack is 390 W,
corresponding to about 1.3 kW/cm3 volumetric heat dissipation. A temperature
gradient from maximal junction temperature to liquid inlet was 60 K. The compu-
tational resource and complexity of detailed computational fluid dynamics (CFD)
and heat transfer modeling in full stacked chips with integrated cooling can be
expensive. Therefore, a porous medium approach was utilized to model the 3D
structure. The model parameters of heat transfer and hydrodynamic resistance are
derived by averaging the results of the detailed 3D-CFD simulations of a single
streamwise row of fins. Variable properties due to temperature gradients are consid-
ered. The porous medium approach with a local thermal nonequilibrium, as well as
orthotropic heat conduction and hydrodynamic resistance, is combined to capture the
heat transfer in the cooling sublayers. The improved porous media model reproduces
the temperature measured in the stack within 10% while lowering the computing
costs for a 3D chip stack nearly 8,000 times compared to those estimated for
modeling the detailed geometry containing 20,000 pins in the chip stack with four
cooling layers.
In 3D chip stacks, the chip architecture design may produce a variety of different
hot spot regions. The hot spot can limit the overall performance of 3D ICs.
The micropin fin array offers the flexibility of various pin diameters for different
heat flux regions. Figure 24 shows the schematic of 3D chip stack package with
integrated cooling, which consists of a row of pins with various diameters and
regions with hot spot (Alfieri et al. 2013). It was found that the local Nusselt number
(Nu) profiles along the channel were affected by the hot spot significantly. The
positive peaks of Nu at the position of the hot spot were observed and revealed a
local heat transfer enhancement, which indicates that the sudden rise in the heat flux
reduces the temperature difference between the liquid and the wall. This effect is
similar to the reason that causes the Nu to be very high at the inlet, where the local
temperature difference between the wall and fluid is very small. For the same reason,
as the heat flux decreases from high to low, the Nu also decreases. The results also
show that at the interface between hot spot and background heating, the local Nu is
even lower than that of only a background heat flux. As the flow develops at the new
imposed conditions, Nu then slowly approaches the corresponding fully developed
value again. Since the hot spot can enhance the Nu, the total heat transfer can be
improved by placing the hot spot near the outlet where the Nu for background heat
flux is usually low due to the fully developed flow. On the other hand, one might
28 Y. Joshi and Z. Wan

Fig. 24 Schematic of integrated liquid cooling of 3D ICs with hot spots

choose to place the hot spot near the inlet to take the advantage of high Nu and thus
limit the hot spot temperature.
The effects of nonuniform pin diameter are also studied. By placing the larger pin
at the hot spot region, the heat transfer area is increased, and also the flow around the
pin is accelerated. Result shows that increasing the pin diameter 2
at the positions
where the hot spot was operating resulted in a rise in the Nusselt number of up to
~30%, depending on the chosen configuration. 3D chip stack can also mitigate the
hot spots on individual dies by strong interlayer heat conduction if the relative
position of the hot spots is selected carefully to result in a heat load and flow
which are well balanced laterally (Madhour et al. 2014).
To manage hot spots, several techniques have been proposed: (1) nonuniform
distribution of microchannels (Shi et al. 2011); more channels were placed above the
hot spot area, while fewer channels above the low-power area. In this case, the
pumping power efficiency can be increased. (2) Workload dynamic management
(Coskun et al. 2009), both reactive and proactive methods, can be employed. In fixed
reactive, the workload scheduler directs the incoming job to the coolest core on the
die. On the other hand, fixed proactive method forecasts future temperature to project
hot core and cool core and, then based on this moves workload from projected hotter
cores to cores that are projected to be cooler. Due to the delay of temperature
response, proactive management can reduce and balance the temperature on the
die more effectively compared to fixed reactive. (3) Dynamic flow rate control
(Coskun et al. 2009); closed-loop control was used to adjust the coolant flow rate
to meet the temperature characteristics of the system. The flow rate can be adjusted
based on the maximum temperature observed during the last measurement interval
or based on the forecast.
Heat removal and power delivery are two major design challenges in 3D stacked
IC (Lee and Lim 2008). These thermal and power/ground interconnects, together
with those used for signal delivery, compete for routing resources, including various
Single- and Multiphase Flow for Electronic Cooling 29

types of through-silicon vias (TSVs). The routing with various interconnects in 3D

was studied. By increasing the microfluidic channel (MFC) depth, mass flow rate
and thus cooling capability can be increased. However, the die thickness is also
increased, and with the fixed aspect ratio of TSVs, the diameter of signal TSVs gets
increased proportionally. Thus, the routing capacity between dies, aggravating signal
net routability, is decreased. Increasing the MFC width also improves mass flow rate
and cooling capability, but the space available for signal TSVs will be decreased.
By increasing the MFC pitch, total heat transfer area between the channel wall and
the coolant will decrease, and thus the cooling performance is degraded, but the
routability of TSVs will be improved.
The thickness of integrated microchannel can affect the performance of the
electrical TSVs (Zhang et al. 2011a). The TSV capacitance increases linearly with
the thickness of the die. For example, assuming aspect ratio limited to 20:1, Ctsv for
a 200 μm micropin fin design is 272.6 fF, while that of a 400 μm micropin fin design,
as optimized, is 847.2 fF. If the TSV aspect ratio keeps increasing, the TSV
capacitance continuously decreases. A heat sink design without consideration of
TSV performance can greatly diminish the advantages of 3D ICs.
Leakage power of the transistors amounts to be a significant part of the total
power consumption but it does not contribute to the computing. The leakage power
has been measured and compared under three different cooling conditions: natural
air cooling, forced convection cooling, and liquid cooling (Wan et al. 2014b). The
liquid cooling is proved to be effective to reduce the leakage power significantly.
Wan et al. (2014a) built a simulation framework which can simulate the power map
and temperature map simultaneously under real applications. Their modeling results
also confirmed that the leakage power needs to be included in the simulation.

4 Flow Boiling in Microgaps

A key limitation of single-phase cooling is the bulk fluid temperature rise along the
flow direction due to the sensible heating, which results in temperature non-
uniformity on the chip. Flow boiling is an alternative thermal management approach
for which the bulk fluid temperature depends on the saturation pressure. It can
achieve higher heat transfer coefficients than in single-phase flow, so reduced fluid
flow rates may be required (Bakir and Meindl 2008). The resulting smaller pressure
drop can result in higher surface temperature uniformity. Two-phase microgap heat
sinks that utilize arrays of microsize pin fins as internal heat transfer enhancement
structures have recently emerged as a promising alternative to the popular two-phase
microchannel array heat sinks to meet the future high heat flux electronic cooling
needs (Wan et al. 2014a). This section will briefly introduce the mechanism of flow
boiling in microchannel, with a focus on flow boiling, and the various flow boiling
enhancement techniques.
30 Y. Joshi and Z. Wan

4.1 Characteristics of Pure Fluid Flow Boiling

The regimes of nucleate boiling and two-phase forced convection govern saturated
flow boiling heat transfer (Qu and Siu-Ho 2009). Nucleate boiling is associated with
bubbly and slug flow. In this regime, liquid near the heated surface is superheated to
a sufficient degree to sustain nucleation. The heat transfer coefficient is dependent
upon heat flux, but fairly independent of mass velocity and vapor quality. The
general trend is increasing h with increasing heat flux due to intensification of
nucleation. The regime dominated by two-phase forced convection, on the other
hand, is often associated with annular flow. In this regime, nucleation is suppressed
along the heated surface, and heat is transferred mainly by conduction across the
liquid film and evaporation at the liquid-vapor interface. As such, h is dependent
upon mass velocity and vapor quality, but less sensitive to heat flux. The general
trend is increasing h with increasing mass velocity and vapor quality due to reduc-
tion in liquid film thickness along the heated surface.
One big disadvantage of flow boiling compared to single-phase liquid cooling is
the instability (Muwanga et al. 2007). Qu and Mudawar conducted flow boiling
experiments with two Si microchannels (Qu and Mudawar 2004). A high-speed
camera was used to determine dominant flow regimes and characterize hydrody-
namic instabilities. When gradually increasing the heat dissipation to the incipient
boiling heat flux, few nucleation sites were observed within the microchannels.
When further increasing the heater power, the flow abruptly transited to intermittent
two-phase flow. At this regime, the boiling two-phase boundary, interface between
the single-phase liquid region and two-phase mixture region, of all the channels
oscillated back and forth between the channel inlet and outlet. For some severe cases,
this oscillation can even result in vapor entering the inlet plenum. During the
oscillation, the inlet and exit pressures, as well as the heat sink temperatures,
fluctuated significantly, and the pump could not deliver a constant flow rate. This
instability was classified as pressure drop oscillation, which is very undesirable for
two-phase microchannel heat sink operation, since it not only leads to large ampli-
tude of pressure and temperature fluctuations but also reduces critical heat flux
(CHF). Premature CHF appeared at a much lower heat flux for instable two-phase
boiling compared to the CHF at steady flow boiling. Fortunately, their experiments
proved the pressure drop oscillation can be suppressed by throttling the control valve
situated immediately upstream of the test module, which helped increase the sys-
tem’s stiffness (Qu and Mudawar 2004).
Another issue with flow boiling is the nonuniform heating (Bogojevic et al.
2011). The boiling usually starts first at high heat flux area. Since the flow resistance
of the vapor region is much higher than that of single-phase liquid region, the fluid
flow will bypass the high heat flux region, which further deteriorates the cooling of
the high heat flux region. The location of the hot spot affects the flow pattern,
pressure, and temperature distribution. When the power map changed from one
pattern to another, the flow boiling pattern, pressure, and temperature also changed,
which further increased the instability of flow boiling.
Single- and Multiphase Flow for Electronic Cooling 31

To prevent the instability, including a throttling valve and inserting an orifice

upstream of the microchannel are proposed. The general idea is to produce a high
pressure drop immediately upstream of the microchannel. This increases the stiffness
of the system and damps the instability. The validity of inlet orifices on saturated
CHF in multi-microchannel heat sinks was studied (Park et al. 2009). Two different
multi-microchannel heat sinks made in copper were tested: one had 20 parallel
rectangular microchannels of 467
4,052 μm (width
depth), and the other had
29 channels of 199
756 μm (width
depth). Three low pressure refrigerants
(R134a, R236fa, R245fa) were used as the coolant. Flow visualization was
conducted with and without an orifice insert at the inlet of microchannels, which
demonstrated the existence of flow fluctuation, backflow, and nonuniform distribu-
tion of flow among the channels when there is no orifice insert. When the boiling
starts and bubbles are generated and grow, they occupy the microchannel, and the
pressure inside the channel increases quickly and pushes the adjacent fluid toward
both downstream and upstream. If there is not a high pressure gradient at the inlet of
the channels, the flow may easily go back into the inlet plenum. Afterward, the
bubbles are observed, to be merged in the inlet plenum, and broke up again to flow
back into the channels.
With the inlet orifice, instabilities such as backflow and thus premature CHF were
successfully prevented. But the inlet orifice affected the flow behavior inside the
channels by creating a jet flow. When boiling starts, the bubbles generated grow and
push the neighboring fluid. While a part of the flow goes downstream, another part of
fluid goes back toward the upstream of the channel. The inlet orifice will block the
backflow due to its high pressure gradient and thus the backflow will merge with the
continuing incoming flow due to the vapor generation. At the same time, the jet flow
from the inlet orifice encounters the backflow, forming a recirculation loop, which
was quite stable and did not create flow maldistribution between channels or
premature CHF. When continue to increase the heat flux, the recirculated bubbles
at the channel inlets started to merge into elongated bubbles. The liquid from the
inlet orifice is injected to the vapor-trapping zone near the channel inlets and broke
up downstream and merged with the liquid-vapor mixture in the boiling region. If
inlet flow has a very low subcooling, the pressure of the fluid drops significantly
across the orifice so that it can be partially flashed. In this case, the liquid entering the
channel will be in the form of a two-phase mixture. In general, without an inlet
orifice, the fluid needs to be superheated over its local saturated temperature to
initiate boiling. With the flashing effect induced by the orifice, bubbles were already
generated in the incoming flow when they enter the microchannel. Therefore,
two-phase flow can be observed even in the unheated entrance region of the
channels. As the orifice results into an additional pressure drop, subcooling of the
fluid is reduced at the actual inlet of the channel. Therefore, boiling starts earlier at a
lower heat flux. This has an advantage of reducing wall-temperature overshoot and
preventing instability. Without the inlet orifices, the wall temperature increased to
values over 100  C at heat flux of 140 W/cm2. However, in the case with the orifices,
it was only 60  C. The flow boiling curves suggest that the orifices can effectively
suppress instabilities and maldistributions.
32 Y. Joshi and Z. Wan

Fig. 25 Tapered manifold design with open microchannels

Fig. 26 Pressure drop

fluctuations over a 20-s period
at different heat fluxes for
uniform (UM) and tapered
(TM) manifolds,
S = 0.127 mm and
V = 80 mL/min (Adapted
from Kandlikar et al. 2013)

Instead of producing a high pressure drop upstream of the microchannel, the

suppression of instabilities can also be achieved by reducing the downstream flow
resistance (Kandlikar et al. 2013). A new tapered manifold design with open
microchannels has been proposed (Fig. 25). The microchannels have an open area
above them with a gap, which provides increased flow area for the fluid. The gap
increases downstream, decreasing the slope of the pressure gradient along the flow
channel and increasing the flow stability.
The effect of the manifold taper on flow stability was evaluated by comparing the
inlet pressure fluctuations at steady-state conditions for both uniform manifold
(UM) and tapered manifold (TM) blocks, with a plain chip and an entry spacing of
S = 0.127 mm. For each heat flux, the pressure drop data over a 20-s period at 5-Hz
frequency are shown. The UM and TM tests were run with a flow rate of V = 80 mL/
min. UM shows pressure fluctuations between 80 and 160 kPa, as seen in Fig. 26 for
heat fluxes in the range of 50–250 W/cm2. TM in the same heat flux range shows
pressure fluctuations below 20 kPa. This indicates that TM can effectively suppress
the flow boiling instability experienced with the UM.
The TM and open microchannel design also helps removal of the vapor from the
heat flux area and supply the liquid to the nucleation sites. The vapor generated in the
microchannel can be removed easily by the manifold above the microchannel since
Single- and Multiphase Flow for Electronic Cooling 33

Fig. 27 Comparison of
boiling performance for the
tapered (TM) and uniform
(UM) manifold blocks with
V = 40 mL/min and
S = 0.127 mm (Adapted from
Kandlikar et al. (2013))

extra space downstream is available for the vapor to flow away from the boiling
surface. Liquid flow is favored inside the microchannel region due to capillary forces
and thus dryout is delayed.
Figure 27 shows the results of the heat transfer performance for the TM and the
UM. The tests were conducted with two different volumetric flow rate: 225 and
40 mL/min. The depth of the tapered manifold is increased by 0.18 mm over the
length of the microchannel region. At V = 40 mL/min, the TM reached 315 W/cm2
at ΔT_sat = 17.2 C, while the UM achieved 308 W/cm2 at ΔT_sat = 20.5 C. Both
manifolds did not reach the CHF limits. The enhanced heat transfer performance of
the TM is believed to be due to its flow stabilization effect. By suppressing the flow
instability, the coolant has an uninterrupted flow passage and a continuous contact
with the heated region. The experiments results also showed that reducing the
manifold depth may provide additional heat transfer enhancement by increasing
flow velocity and by further preventing backflow.
Surface microstructures in two-phase microchannels can also be modified to
suppress flow instabilities and enhance heat transfer (Zhu et al. 2016). By fabricating
silicon micropillar arrays on the bottom of the microchannel wall (Fig. 28), capillary
flow for thin film evaporation can be enhanced, while facilitating nucleation only
from the side walls.
The single microchannel has dimensions of 10 mm in length, 500 μm in width,
and 500 μm in height. The micropillar on the bottom surface of the microchannel has
diameters of 5–10 μm, pitches of 10–40 μm, and heights of 25 μm. These designs of
the micropillar geometries were based on the following reasons: (1) the fabrication
of the microchannel is easy and controllable using the standard Si etching process,
(2) the capillary pressure is high enough to manipulate the flow behavior, and (3) the
surface structures are mechanically strong and will not break or deform as the liquid
evaporates.
34 Y. Joshi and Z. Wan

Fig. 28 Microchannel with micropillars

Performance between the smooth surface and structured microchannels at mass

flux 300 kg/m2 s was compared. At low heat flux q00 < 400 W/cm2, both smooth
surface and the structure surface microchannels show similar and small fluctuations.
When the heat flux increased to q00 = 430 W/cm2, the smooth surface microchannels
show temperature spikes and increased pressure drop fluctuations. However, small
temperature (5  C) and pressure drop fluctuations (300 Pa) are observed for all
the structured surface microchannels. When continuing to increase the heat flux, the
smooth surface microchannel shows large amplitude (>20  C) periodic dryout,
while the structured surface microchannel showed stable temperature and pressure
drop at the same heat flux. The smooth surface could not supply liquid to the surface
and maintain this liquid film and thus the dryout area expanded. For the structured
surface, wicking capability of the microstructures helps wet the surface and maintain
the liquid film. Visualization of the flow pattern and the dryout process demonstrates
that the micropillar surface can enhance capillary flow and increase flow stability by
maintaining a stable annular flow and thin film evaporation. The structured surface
microchannel contributes to an enhanced heat transfer coefficient and CHF (maxi-
mum 57%) compared to a smooth surface microchannel. In addition, the additional
pressure drop introduced by the micropillar structures was negligible. Silicon nano-
wire coating has also been investigated to suppress flow instability and enhance heat
transfer (Li et al. 2012).
Figure 29 shows the boiling curve for the plain surface and nanowire-coated
surface. At single-phase flow region for both mass fluxes, the heat transfer charac-
teristics are almost the same for both plain-surface and nanowire-coated micro-
channels. However, for the latter, an earlier ONB is observed at a lower heat flux.
The visualization study revealed abundant and much more uniformly distributed
bubbles in the nanowire-coated microchannels. Similar to pool boiling, this was
attributed to the larger number of nucleation sites in the nanowire-coated micro-
channel. After ONB, the onset of flow oscillation (OFO) was immediately observed,
which resulted in a very limited stable flow boiling condition for plain-surface
microchannels. On the other hand, stable flow boiling was observed for a much
larger range of heat fluxes for nanowire-coated microchannels, which indicates that
nanowire-coated microchannels could significantly suppress flow instability and
delay the OFO. Figure 29 also shows that the nanowire-coated microchannels can
significantly enhance heat transfer coefficient compared to plain microchannels.
Such enhancement can be attributed to the unique properties of nanowire coatings,
Single- and Multiphase Flow for Electronic Cooling 35

Fig. 29 Flow boiling curves

for plain-surface and
nanowire-coated
microchannels for mass fluxes
at G = 476 kg/m2 s. Tests are
conducted at Tin = 60  C
(Adapted from Li et al.
(2012))

including high nucleation site density, large surface area, large capillary force, and
superhydrophilicity.
The previous stabilization methods are passive but come usually at the cost of
high pressure drop penalty. Active control methods can suppress inherent flow
instabilities, especially in transient applications, thus improve the coefficient of
performance. Based on the theoretical analysis of oscillatory flow boiling, a set of
active control schemes are developed and studied to suppress flow oscillations and to
increase the CHF. With the available control devices – inlet valve and supply pump –
different active control schemes are studied to improve the transient two-phase
cooling performance (Zhang et al. 2011b). For effective transient electronic cooling,
the main control objectives are to maintain high heat transfer performance of
two-phase flow inside the heated channel and to avoid pressure drop and flow
oscillations, even under large transient heat load changes. Two control elements
are available for feedback: the inlet valve of the boiling channel and the supply pump
in the system. Inlet valve-based control methods are widely used in practice. In the
pump-based flow control system, the manipulated variable – inlet flow rate – is
linearly dependent on the positive displacement pump voltage. Therefore, inlet flow
rate is not fixed but changes with heat load, and the change rate can affect the flow
behavior of both the tank and the boiling channel.
Inlet valve-driven feedforward control: a control valve before the heated channel
can be used to suppress flow boiling instability. However, to control valve induces
higher pressure loss and potentially higher supply pumping power compared to a
system without an inlet valve. So active valve control strategies instead of fixed inlet
valve for the suppression of two-phase flow instabilities are more desirable. The
most straightforward method is to keep the valve fully open when the superheated or
subcooled flow is stable, while reducing the valve opening position when the flow is
not stable. Modeling results of transient control responses showed that the
36 Y. Joshi and Z. Wan

oscillatory behavior of mass flux and pressure is eliminated with the valve controller
even when the heat load changes.
Supply pump-driven feedback control: although the inlet valve suppresses the
upstream compressible flow instability, it increases pressure loss and pumping
power. Alternatively, the inlet-positive displacement pump can be used to regulate
the downstream flow conditions while removing the inlet restrictor (valve) so that no
additional pressure loss is induced. It was found that with the pump-driven active
feedback control scheme, the flow oscillation can be successfully suppressed. The
controller is implemented. The inlet refrigerant flow is not kept constant; instead, it
will change with transient flow and heat load conditions.
One important parameter with flow boiling is the critical heat flux (CHF), which
determines the maximum heat flux the microchannel can dissipate before the device
fails. After CHF is reached, a stable liquid film cannot be formed, vapor occupies the
channel, and heat is transferred to the vapor mainly by conduction. So, one method
to increase the CHF is to remove the vapor immediately after it is generated. A novel
two-phase microchannel cooling device that incorporates perforated side walls
(Fig. 30) was proposed for potential use as an embedded thermal management
solution for high heat flux semiconductor devices (Warrier et al. 2014). In this
cooling device, perforated side walls are used to form the channels. Each liquid
channel is separated from the two adjacent vapor channels. While flowing in the
liquid channels, the liquid is transported through the perforations by capillary force
and evaporates into the vapor channels. In this case, both the liquid and vapor flow in
the single-phase regime. Evaporation only occurs in the arrays of micro-perforations
in the side walls of high thermal conductivity. This design enables one to circumvent
flow instability and dryout. The unique perforated channel walls of high aspect ratio
have two benefits: it can serve as heat transfer surface and separate the vapor and
liquid. Thermo-fluid modeling results for one specific implementation of the device
concept predicted pressure drop of <1,000 Pa along 2.5-mm-long main channel
sections for FC-72 and almost 2
reduction in the peak temperature rise at a chip
heat flux of 1,000 W/cm2, when compared with single-phase microchannel flows at
even higher pressure drops.
Surface characteristics such as wettability and nucleation site density of micro-
channels are very important parameters for CHF. Usually, high wettability and
nucleation density are preferred. Flow boiling cycle includes bubble nucleation,
growth, separation, interactions, development of two-phase flow regimes, and
rewetting which is greatly influenced by surface properties. A number of studies
have been performed to enhance heat transfer in microchannels by controlling
bubble nucleation site density and wettability via modifying surface properties.
The heat transfer enhancement attributes of carbon nanotubes (CNTs) applied to
the bottom wall of a shallow rectangular microchannel were examined (Khanikar
et al. 2009). Flow boiling experiments were conducted with both a bare copper
bottom wall and a CNT-coated copper wall with deionized water as working fluid. It
was found that CNT arrays facilitate earlier, abundant, and intense bubble nucleation
at low mass velocities. However, at high mass velocities, enhancement in the
nucleate boiling region is compromised.
Single- and Multiphase Flow for Electronic Cooling 37

Fig. 30 Two-phase microchannel cooling device with perforated side walls

Figure 31 shows the variation of CHF with mass velocity for both the bare and
CNT-coated walls for inlet temperatures of 30  C and 60  C. At the lowest mass
velocity of 86 kg/m2s, the CHF of CNT-coated wall was enhanced by 23.4% at
Tin = 30  C and 21.1% at Tin = 60  C. However, as the mass velocity increases to
G = 268 kg/m2s, the enhancement diminished to 3% for Tin = 30  C and 20.8% for
Tin = 60  C. At the highest mass velocity of G = 368 kg/m2 s, the CHF performance
of the CNT-coated wall was further compromised, resulting in a 22.3% reduction in
CHF for Tin = 30  C and only 2.8% enhancement for Tin = 60  C. Overall, the
CNTs can enhance the CHF at low mass velocities, but the enhancement diminishes
at high mass velocities. It is believed that CNTs can maintain a near-vertical
orientation and protrude into the liquid flow at low mass velocities and thus increase
both wall roughness and turbulence. However, at high mass velocities, CNTs tend to
deform and fold upon the heated wall, decreasing the wall roughness and turbulence.
In terms of boiling incipience, CNT arrays appear to be effective at both initiating
and sustaining the nucleation process due to the modified vapor embryo entrapment.
Silicon tips with nanoscale scallops (Fig. 32) in Si microchannels can be easily
fabricated by Si etching method (Yang et al. 2015). These silicon tip arrays can
significantly increase the nucleation site density. As a result, flow boiling heat
transfer of a dielectric refrigerant, HFE-7000, is found to be substantially enhanced
in a microchannel heat sink. Compared to plain-wall microchannels, the average
junction temperature can be reduced up to 10  C at a heat flux of 55 W/cm2, and the
equivalent thermal resistance is reduced up to 31% at a mass flux of 1,018 kg/m2s.
At a lower mass flux of 1,018 kg/m2 s, and moderate heat flux from 40 to 60 W/cm2,
boiling heat transfer is dominated by nucleate boiling. The microscale gaps formed
between Si tips and nanoscale scallops on Si tips can serve as potential bubble
nucleation sites to enhance nucleate boiling. However, at moderate mass flux of
1,527 kg/m2 s, the temperature reduction is less than 6.5  C. It should be noted that
the introduction of Si tip arrays actually decreases CHF. It is believed that the
capillary force is not strong enough to drive dielectric liquid flow, since the surface
38 Y. Joshi and Z. Wan

Fig. 31 Variation of CHF

with mass velocity for bare
and CNT-coated walls for
inlet temperatures of 30  C
and 60  C (Adapted from
Khanikar et al. (2009))

Fig. 32 Silicon tips in Si

microchannels

tension for HFE-7000 is much smaller. The Si tips, if not helping liquid rewetting,
would add additional flow resistance to liquid flow close to walls and may induce
premature CHF.
Copper nanowire (CuNW) coating can significantly increase the wettability of
microchannel surface (Morshed et al. 2012). The contact angle is decreased from
97.1 to 57.6 with the CuNWs. Subcooled flow boiling for water at different mass
velocity (45.9–143.8 kg/m2s) and a constant inlet temperature of 22  C are studied
and compared between CuNW-coated microchannel and bare surface microchannel.
A boiling curve was created. A significant change in the slope of the boiling curves
can distinguish the single-phase and two-phase zones. In the single-phase zone,
surface temperature increases almost linearly with heat flux and also decreases with
Single- and Multiphase Flow for Electronic Cooling 39

the mass flow rate. For all the cases studied, heat transfer rate is clearly higher for
CuNW-coated surface. In single-phase region, heat transfer coefficient was indepen-
dent of heat flux both for CuNW-coated surface and bare surface, but dependent on
mass velocity as expected. CuNW coatings enhance single-phase heat transfer rate
by up to ~25% compared to the bare surface. In the flow boiling regime, the
enhancement was up to ~56% with a pressure drop increase of ~20% compared to
the single-phase regime. Nanowire coating wetted surface area, surface roughness
(fin effect), and surface wettability increase are all possible factors for this heat
transfer coefficient enhancement in the single phase. Once the boiling starts, slope of
the boiling curve is observed to change significantly. With the increase of mass
velocity, the single-phase flow suppresses bubble nucleation, resulting in delayed
boiling incipient point. CuNW-coated surface reduces the surface superheat temper-
ature required for incipience of boiling significantly.
A hybrid microchannel/slot-jet module that combines the cooling attributes of
microchannel flow and jet impingement presented by Sung and Madawar (2008)
shows high heat remove ability in single-phase cooling. The subcooled nucleate
boiling and critical heat flux (CHF) characteristics of such hybrid cooling module
were also evaluated (Sung and Mudawar 2009) using HFE-7100 as working fluid.
Increasing the coolant’s flow rate and/or subcooling shifted both the onset of boiling
(ONB) and CHF to higher heat fluxes and higher wall temperatures. The hybrid
module achieved heat fluxes as high as 1,127 W/cm2, reported to be the highest
value ever achieved for a dielectric coolant at near atmospheric pressure. The hybrid
cooling configuration involves complex flow interactions between circular jets and
microchannel flow. A homogeneous model was developed to capture the fluid flow.
CHF is obtained by a superpositioning technique that divides the heated wall into
two portions, one dominated by jet impingement and the other microchannel flow.
This technique is shown to predict the CHF for the hybrid cooling module with high
accuracy.
The hybrid cooling module uses microjet to inject the fluid from the top of
microchannel and thus increase the flow mixing. However, this configuration
might not be suitable for 3D IC cooling. A modified design has been proposed: a
microchannel array consisting of each main channel connected to two auxiliary
channels (Dai et al. 2013), as shown in Fig. 33. By routing a portion of the incoming
flow through a passive microjet, the boundary layer growth can be break up in the
single-phase flow. For the flow boiling, as bubbles generate in the main channel and
move to the microjet, the subcooled liquid from the microjet can break up the vapor
and thus enhancing the flow mixing. This arrangement was shown to effectively
suppress bubble growth and flow instability, as well as introduce mixing, and
manage the bubble expansion rate in the two-phase regime. These effects contributed
to a more uniform temperature distribution and heat transfer performance enhance-
ment. Specifically, compared with a conventional microgap under similar working
conditions, the auxiliary channel achieved a reduction of average wall temperature
for the convective boiling by 2.9  0.6  C at a mass flux of 83 kg/m2s with a 60.4%
reduction in the pressure drop. Moreover, at a mass flux of 373.5 kg/m2s, CHF
reached 311 W/cm2 in the two-phase regime. The reduction of the pressure drop is
40 Y. Joshi and Z. Wan

Fig. 33 Concept of the flow separation technique. (a) Fluid flow in the microgap during single-
phase flows. (b) Fluid flow in the microgap during flow boiling

mainly due to the additional auxiliary channel and the effective suppression of the
bubble expansion. However, the surface temperature reduction was shown to
decrease with increasing mass flux. This could be due to the reduced flow rate
through the auxiliary channel.
The previous two-nozzle configuration confines the two-phase oscillations down-
stream of the microchannels. A four-nozzle microchannel configuration was also
proposed to further enhance mixing to the entire channel and thus increase the heat
transfer performance (Li et al. 2016). One auxiliary channel was shared by two
adjacent main channels to achieve a denser channel array, and four micronozzles are
uniformly distributed along the side wall (Fig. 34). Flow boiling experiments in the
four-nozzle microchannel was conducted with deionized water as the coolant over a
mass flux range of 120–600 kg/m2 s. Compared with the previous two-nozzle
design, four-nozzle configuration not only enhances the mixing to the entire channel
but also increased nucleation sites, which was confirmed by a visualization study.
Overall, four-nozzle configuration enhanced the average heat transfer coefficient
approximately by 54.5% compared to the two-nozzle configuration. Simultaneously,
the pressure drop was further reduced by approximately 50%. More uniform wall
temperature and stable flow were achieved with this configuration.
Compared to the extensive study of microchannels, only a few studies have been
done on flow boiling of micropin fin array-enhanced microgaps. Qu and Siu-Ho
(2009) conducted experiments to measure heat transfer coefficient for saturated flow
boiling of water in an array of staggered square pin fin microgap. For a near-
saturated inlet condition, h was fairly constant in the high-quality region, insensitive
to both quality and mass velocity. Heat transfer in the low quality region was
enhanced by inlet subcooling. The enhancement effect due to the inlet subcooling
diminished with increasing quality. Two-phase forced convection associated with
annular flow was postulated to be the governing heat transfer mechanism.
Krishnamurthy and Peles (2008) studied the flow boiling of water across a bank of
Single- and Multiphase Flow for Electronic Cooling 41

Fig. 34 (a) Two-nozzle microchannel. (b) Four-nozzle microchannel

circular staggered micropin fins, 250 μm long and 100 μm diameter, with pitch-to-
diameter ratio of 1.5. For mass flux G = 364 kg/m2 s at heat fluxes slightly above the
onset of boiling, vapor formed at nucleation sites in the vicinity of the channel exit.
The hydraulic resistance imposed by the pin fins forced the bubbles to oscillate
between the pin fins. The sliding motion of the bubbles caused the thin liquid film
between the surface and the bubble to evaporate and to form vapor cavity. High
surface tension forces acting along the vapor-liquid interface, low superficial liquid
velocity, and hydraulic resistance of the pin fins caused these vapor cavities to
remain stationary. With increasing heat flux, the vapor cavities grew, resulting in
an increase in the local void fraction. This in turn increased the local superficial
liquid velocity and, thus, the drag force on the vapor cavity forcing the bubbles to
propagate downstream. At higher heat fluxes, annular flow was established. The heat
transfer coefficient depended moderately on the mass flux and weakly on the heat
flux. This trend suggested that the dominant heat transfer mechanism was convective
boiling, which was further supported by the presence of annular flow observed
during flow visualization. This convective flow boiling was also observed by Isaacs
et al. (2012, 2013). However, an experiment by Reeser et al. (2014) found that the
average heat transfer coefficient appeared almost independent of mass flux and
increased with heat flux.
42 Y. Joshi and Z. Wan

Due to the microscale flow passages and vapor production inside the heat sink,
excessive pressure loss is always a concern, which can lead to elevated pumping
power consumption and high operating cost. Konishi et al. (2009) found that for a
given mass flux, pressure drop increased rapidly with increasing exit vapor quality
xe. For a fixed xe, pressure drop increased appreciably with increasing mass flux.
Heat transfer coefficient was found to decrease sharply at low exit quality and low
heat flux and then plateaued at moderate to high exit qualities and heat fluxes
(Qu and Siu-Ho 2009). However, Reeser et al. (2014) showed that the average
heat transfer coefficient increased monotonically, as xe increased from 0 to 25%. It
was found that the in-line and staggered data points nearly coincided over the entire
range of qualities, implying that – contrary to conventional wisdom – neither the
in-line nor staggered array was significantly better than the other in terms of cooling
performance in this parametric range.
The selection criteria for coolant for two-phase cooling are different from single-
phase cooling due to different working conditions. For example, operating pressure
is a major consideration for two-phase cooling. The saturation temperature of water
is 100  C at atmosphere pressure, which is too high for CMOS chips to effectively
utilize the benefit of latent heat. Although water is the best coolant, its high electrical
conductivity makes it undesirable to use for direct chip-level cooling. Dielectric
fluids such as Fluorinert and refrigerants are electrically nonconductive and therefore
preferred over water for direct contact cooling over long usage durations from
materials’ compatibility and reliability perspective for flow loops comprised of
components such as metal tubing. R-123 was used as the fluid to study the boiling
inception and pressure drop over a bank of micropin fin arrays 243 micrometer long
with hydraulic diameter of 99.5 μm (Kosar and Peles 2006). Boiling initiation was
suppressed to high liquid superheat temperatures due to the hydrophilicity of the
working fluid on silicon surfaces. R-123 has a near-zero contact angle on silicon, and
as a result, large nucleation cavities were flooded, which in turn gave rise to high
liquid superheat at boiling inception. The delay in boiling to high superheats
triggered a chain of periodic events. Once boiling was initiated, vapor burst insta-
bilities were noticeable.
McNeil et al. (2010) compared flow boiling heat transfer in in-line pin fin and
plane channels using R113 at atmospheric pressure. The mass flux range was
50–250 kg/m2s, and the heat flux range was 5–140 kW/m2. The measured heat
transfer coefficients for the pin fin surface were slightly higher than those for the
plate surface. Both were dependent on heat flux and reasonably independent of mass
flux and vapor quality. Thus the heat transfer was probably dominated by nucleate
boiling. The pin fin pressure drops were typically seven times larger than the plate
values.
HFE-7200 has also been studied as a dielectric coolant (Reeser et al. 2014).
Unlike the observed water behavior, the HFE-7200 data displayed an approximately
50% improvement in the average coefficient for the staggered array over the in-line
array, for much of the range of exit qualities. Most notable for both HFE-7200 array
configurations, however, was the initial sharp decline in the average heat transfer
coefficient from the lowest exit qualities to about 10–15%, followed by a plateauing
Single- and Multiphase Flow for Electronic Cooling 43

or mild increase up to exit qualities of 40–50%, where it reached a local maximum.

Finally, the average heat transfer coefficient deteriorated, as the exit quality
approached 100%, possibly reflecting localized dryout in the pin fin arrays.
Ma et al. (2009) studied the flow boiling heat transfer performance of FC-72 on
silicon chips with micropin finned surface. Channel liquid velocities of 0.5, 1 and
2 m/s were tested at inlet subcoolings of 15, 25, and 35 K. The micropin finned
surface had considerable heat transfer enhancement compared to a smooth surface.
The CHF values for all the surfaces increased with fluid velocity and subcooling.
The effect of fluid velocity was more notable, especially for fluid velocity larger than
1 m/s. The CHF of the micropin fin surface was more sensitive to the fluid velocity
and liquid subcooling than that of the smooth chip. For a lower ratio of pin fin height
to fin pitch, and/or higher fluid velocity, the forced flow and heat transfer on the chip
had a greater effect on the bubble nucleation, and the entire micropin finned surface
was not completely covered with bubbles, creating a dominant convective heat
transfer effect in the nucleate boiling region.

4.2 Flow Boiling Modeling

Flow boiling simulation in microchannels remains a challenging problem, but the

increasing interest in applications of microscale two-phase cooling highlights its
importance. Various models have been proposed to simulate the flow boiling in
microchannel. Generally, there are two kinds of models: homogeneous and inhomo-
geneous. The former treats the mixture of vapor and liquid as a single fluid, with the
fluid properties depending on the quality. The resulting mass, momentum, and
energy conservation equations for the fluid are solved:

@ρh 1 @Acr ρh j
þ ¼0 (2)
@t Acr @s

@ρh j 1 @Acr ρh j2 @P
þ ¼  gρh sin θ  Fw (3)
@t Acr @s @s
@ρh e 1 @Acr ρh je @P q00w Pm
þ ¼ þ þ Fw j (4)
@t Acr @s @t A
The homogeneous model divides the flow into three regions for the convective
heat transfer coefficient (Saenen and Thome 2015): the single-phase convection, a
small subcooled boiling region, and the two-phase saturated flow boiling region. The
vapor quality is used to determine the flow regime transition 0.01 for transition
from single-phase liquid to subcooled boiling and 0 for subsequently transitions to
saturated flow boiling. The transition from two-phase saturated flow boiling region
to the single-phase vapor convection occurs at either the critical vapor quality,
indicating dryout, calculated using the CHF correlation, or at a vapor quality of
one, whichever occurs first. Since the homogeneous model does not track the vapor-
44 Y. Joshi and Z. Wan

Fig. 35 Porous media model of flow boiling in microchannel

liquid interface, and relies largely on empirical correlations, it enables fast dynamic
modeling of flow boiling at less computational cost.
For modeling of a full chip, the 3D-CFD modeling requires large number of
meshes and computation time. Due to this difficulty, porous media method has been
introduced for effective modeling (Zhang et al. 2013). Homogeneous flow boiling
model can be combined with porous media method to further facilitate modeling of a
large number of microchannels. Thermal fluid transport in the microchannel array is
represented as a porous media (Fig. 35) governed by empirical correlations. Com-
pared with the combined conduction and flow modeling, which considers individual
channels and walls, this modeling method only considers one porous medium layer,
instead of many individual channels, and uses empirical correlations instead of the
complex two-phase CFD formulations. As a result, it allows for rapid exploration of
thermal design alternatives. The effects of channel width variation on two-phase
cooling performance were analyzed. The adjustment of channel widths on a chip can
reduce the maldistribution of flow rates and exit qualities among the channels and
improve the temperature uniformity on the chip.
The inhomogeneous model treats the liquid and vapor as two immiscible fluids,
and the interface is tracked during the simulation. Volume of fluid (VOF) model is a
common homogeneous model for detailed flow boiling simulation (Lorenzini and
Joshi 2015). The superficial velocity of the liquid and vapor is assumed to be the
same. The volume fraction for each phase in every control volume is calculated, and
the interface is tracked by the volume fraction. A single set of momentum and energy
conservation equations is shared between the phases. The interface mass transfer due
to the vaporization/condensation is estimated based on the mechanistic model
proposed by Lee (1980):

T l  T sat
Sl ¼ λl αl ρl ; Sv ¼ Sl ; if T l > T sat ðvaporizationÞ (5)
T sat
T sat  T v
Sl ¼ λv αv ρv ; Sv ¼ Sl ; if T v > T sat ðcondensationÞ (6)
T sat
Single- and Multiphase Flow for Electronic Cooling 45

Fig. 36 Schematic of the two-zone model

VOF methods for flow boiling simulation are promising due to the intrinsic mass
conservation achieved, unlike with other Eulerian methods. However, VOF uses
fixed-grid approaches to estimate the interface profile, which may lead to unphysical
flows due to the numerical error in estimating the interface surface tension. In order
to suppress the unphysical flow, evaporative heat and mass source terms are calcu-
lated using a saturated-interface-volume phase change model which fixes the inter-
face at the saturation temperature at each time step to achieve stability (Pan et al.
2015). This modified model can eliminate numerical oscillation of the evaporation
source terms. With the help of a non-iterative time advancement scheme, the
computational cost is reduced. The reference frame is set to move with the vapor
slug to artificially increase the local velocity magnitude, which reduces the influence
of numerical errors from calculation of the surface tension force, and thus suppress
the development of spurious currents. So nonuniform meshes can be used to
efficiently resolve high aspect ratio geometries and flow features and significantly
reduce the overall numerical expense. The mass and heat transfer source terms in the
VOF modeling depend on the specific flow boiling regime, since the interface
dynamics and heat transfer depend on the specific regime. Harirchian and Garimella
(2012) identified four regimes for flow boiling in microchannel: slug, confined
annular, bubbly, and alternating churn/annular/wispy-annular flow. Separate models
must be used to accurately model the flow boiling in different regimes.
VOF approach with a two-zone model is used to model the slug flow, which is
characterized by elongated bubbles separated by liquid slugs (Magnini and Thome
2015). As shown in Fig. 36, the vapor bubble and liquid slug pair are decomposed
into liquid film, dry vapor plug, and liquid slug zones within this model. In the full
vapor and liquid regions, single-phase convection is assumed to govern heat transfer.
Heat transfer coefficient in these zones is evaluated by single-phase correlations for
hydrodynamically and thermally developing flow. In the liquid film region, steady-
state heat conduction is assumed to dominate heat transfer. The liquid film dryout
process is assumed to be triggered by the local evaporation of liquid, exclusively.
The liquid film thickness varies as it evaporates along the microchannel and is
estimated based on the local bubble velocity and acceleration by implementing a
correlation for circular channels. A submodel which accounts for the actual liquid-
vapor slip was used to calculate the bubble velocity and acceleration. Superheating
46 Y. Joshi and Z. Wan

of the liquid within the slug can also be properly evaluated. The liquid film dryout is
not included in the boiling model as it does not occur under the simulation setup.
In the VOF model, the interface is implicitly captured by the volume fraction in
each control volume. Therefore, it can handle the topology change of the interface.
However, the computation is limited to a fluid with small liquid-vapor density ratio
(Ansys 2016). To model fluid with high liquid-vapor density ratio in microchannel, a
level-set (LS) method can be used (Lee and Son 2008). The liquid-vapor interface is
determined by the LS function, which is defined as a signed distance from the
interface. Since the distance function and its spatial derivatives are smooth and
continuous, the LS method can be used to compute an interfacial curvature more
accurately than the VOF method.
Although the inhomogeneous model provides better spatial and temporal resolu-
tion than homogeneous model, and enables understanding of the detailed bubble
dynamics, the homogeneous model has several advantages. First, it does not need to
solve complex, flow pattern-dependent, interfacial transport terms required for
multi-fluid models. Since these transport terms are largely unknown or not well
validated for microchannel two-phase flow, the added accuracy of using an inhomo-
geneous model would be questionable. Second, accurate or widely applicable void
fraction correlations or models are not yet available for flow boiling of micro-
channel. On the other hand, homogeneous void fraction matches well with many
experimental void fraction measurements in microchannels. Therefore it is uncertain
that using more complex models, such as a drift-flux model, or a diffusion model,
would increase the accuracy (Saenen and Thome 2015). Finally, a typical micro-
channel evaporator consists of 20–100 microchannels. Modeling all these micro-
channels with a 2D or 3D interface tracking model would be computationally very
expensive. Also, the model would need to be able to accurately simulate a wide
variety of flow patterns and conditions in microchannel, which is not straightforward
to accomplish.

4.3 Flow Boiling of Fluid Mixtures

Fluid mixtures have been a research topic for boiling enhancement and studied
extensively. A large body of work has focused on pool boiling of mixtures, and
the enhancement mechanisms have been studied. Van Wijk et al. (1956) studied the
mixtures of water with acetone, alcohols, ethylene glycol, and methyl ethyl ketone.
A CHF enhancement was achieved at an optimum concentration. They concluded
that this enhancement in CHF was due to reduction in the bubble departure diam-
eters. McGillis and Carey (1996) investigated pool boiling of mixtures of water with
ethylene glycol, methanol, and 2-propanol. Addition of small amount of alcohol to
water enhanced the CHF. The mixtures were classified into positive (more volatile
component having lower surface tension) and negative (less volatile component
having lower surface tension) mixtures. Due to the differences in fluid volatility,
preferential evaporation of one component occurred along the liquid-vapor interface
of a binary mixture. The variation in concentration along the liquid-vapor interface
Single- and Multiphase Flow for Electronic Cooling 47

resulted in a surface tension gradient due to the Marangoni effect. If the surface
tension of the more volatile component was less than that of the less volatile
component, the concentration gradient would generate a force that pulled the liquid
toward the heated wall. If the surface tension of the more volatile component was
greater than that of the less volatile component, a force that pulled the liquid away
from the heated surface could be generated. The study by Hovestreijdt (1963) and
Fujita and Bai (1997) attributed the CHF enhancement to the Marangoni effect.
Kandlikar and Alves (1999) performed pool boiling experiments using mixtures of
water with ethylene glycol at low concentrations (1–10 wt. %). The effects of surface
tension gradients were negligible at low mixture concentrations, and they attributed
the observed improvement in heat transfer coefficient to the changes in contact angle
and wetting characteristics of the mixture. Arik and Bar-Cohen (2010) observed
significant CHF enhancement using mixture of FC-72 and FC-40. They attributed
the enhancement to the improvement in thermal properties of the mixture.
There exist a few studies on flow boiling of mixtures. Peng et al. (1996) and Lin
et al. (2011) studied the flow boiling of water-methanol mixtures in microchannels.
The CHF increased at low concentration but decreased as the concentration of
methanol in water increased. The enhancement was attributed to the Marangoni
effect. However, heat transfer degradation was also observed by other studies.
Bennett and Chen (Bennett and Chen 1980) observed a significant reduction in
heat transfer coefficient for mixtures of water and ethylene glycol and attributed it to
mass transfer effects. Kandlikar and Bulut (2003) studied the flow boiling of
ethylene glycol and water. The heat transfer performance deteriorated as ethylene
glycol concentration increased. They also attributed the degradation to the mass
transfer. Sathyanarayana (2013) conducted flow boiling experiment with 20 wt. %
mixture of HFE 7200 – methanol in a microgap channel. The CHF enhancement was
attributed to the smaller bubble departure diameter.

4.4 System Configuration and Design Consideration

System design considerations are drastically different from single-phase flow. The
properties of the coolant are very important for the design of a flow boiling cooling
system. The pressure-dependent saturation temperature determines the operation
temperature of the chip. Surface tension affects CHF and determines the maximum
heat dissipation rate. For example, water has much higher thermal conductivity and
latent heat of vaporization than typical refrigerants and dielectric coolants, such as
R-134a and HFE-7200. However, water has much higher saturation temperature at
atmosphere pressure. Therefore, for electronic cooling, subatmospheric system
pressure is required with water to ensure a sufficiently low operating temperature.
Table 2 shows the thermophysical properties of several selected coolants (Mohapatra
2006 and Ellsworth 2006).
Instability is a significant issue with flow boiling. Fluctuations of flow rate,
pressure drop, and temperature are not desirable and may cause damage or failure
of the device to be cooled. Throttling valve or inserting orifice can be used to
48 Y. Joshi and Z. Wan

Table 2 Thermophysical properties of selected coolants (Mohapatra 2006 and Ellsworth 2006)
Dynamic viscosity Thermal conductivity Specific heat Density
Coolant (kg/m s) (W/mK) (J/kg K) (kg/m3)
FC-77 0.0011 0.06 1,100 1,800
R134a 0.00001 0.014 854 4.23
R245fa 0.00001 0.01405 920.5 5.675
Water 0.00089 0.61 4,180 997
HFE-7200 0.00067 0.068 1,214 1,424

suppress the instability, but at the penalty of higher pressure drop. As the vapor
generates and expands in the microchannel, the system pressure is elevated signif-
icantly. Thus, device reliability and safety are a big concern. Dryout must be avoided
since the associated excessive temperature rise can burn the device.
Both direct and indirect cooling systems (Fig. 37) have been explored (Lee and
Mudawar 2009). In the first direct refrigeration configuration, the cooling module of
the electronic device is incorporated as an evaporator in a vapor compression cycle,
and the refrigerant serves as coolant. The alternative indirect refrigeration cooling
configuration uses two fluid loops. Heat from the device is dissipated to the coolant
in the primary flow loop and then rejected through a heat exchanger to the coolant
flowing in a separate vapor compression cycle. In the direct refrigeration cooling
system, microchannel heat sink serves as an evaporator. So the coolant’s operating
conditions need to conform to those of a vapor compression cycle. First, for most
refrigeration compressors, the refrigerant entering the microchannel evaporator
should be a two-phase mixture and exits as saturated or superheated vapor. Saturated
or superheated vapor conditions are favored for safe operation of the compressor.
This requires the use of a secondary heater or phase separator downstream of the
microchannel. For indirect cooling system, using a separate loop for the primary
coolant enables attaining the desired microchannel heat sink’s inlet conditions. Most
importantly, the coolant is not required to be maintained in a near-saturated or
superheated state, as required by the compressor in the direct cooling configuration.
In the primary loop, the inlet fluid of the heat sink can be highly subcooled.
Compared to saturated boiling, subcooled boiling increases the CHF significantly,
which is beneficial for high heat flux cooling. A secondary flow loop is used to
dissipate heat to the ambient by using a fluid-to-fluid heat exchanger. With the
two-loop structure, distributed multiple heat loads could be handled with simple
and small pump loops, all of which could be coupled to a centralized chiller
(secondary flow loop).
However, highly subcooled boiling of two-phase cooling systems suffers from
various flow boiling instabilities. Flow boiling oscillations can modify the hydrody-
namics of the flow, can generate acoustic noise, and can endanger the structural
integrity of the system. Most importantly, flow instability can lead to premature
initiation of the CHF condition.
Single- and Multiphase Flow for Electronic Cooling 49

Fig. 37 (a) Direct cooling system. (b) Indirect liquid-cooling system

5 Summary

Liquid cooling, as a promising technology, can handle very high heat flux compared
to air-cooled heat sinks. With microfluidic cooling, small high heat flux components,
such as microprocessors, can be kept at lower operating temperature, displaying
improved performance. For large facilities such as data centers, significant amounts
of energy can also be saved due to the leakage power reduction from chips, as well as
the reduction in cooling power. Understanding of characteristics of liquid cooling,
both under single-phase and two-phase conditions, is essential for its successful
applications.
50 Y. Joshi and Z. Wan

In single-phase liquid cooling using microchannel arrays, the heat transfer per-
formance is compromised by saturation of the heat transfer coefficient, and the bulk
fluid temperature rises. Various heat transfer enhancement techniques such as wavy
channels, adding pin fins and incorporating microjet array, have been proposed to
enhance the flow separation and mixing, which can increase heat transfer coefficient.
However, the increase of heat transfer is usually accompanied by increased
pressure drop.
Liquid cooling is an attractive cooling technology for 3D ICs. However, the integra-
tion of 3D ICs with liquid cooling is not well demonstrated and requires the consider-
ation of signal, power, and liquid cooling, since they compete for the available space
and may interact with each other. In addition, not all the liquid-cooling methods are
suitable for integration into 3D IC applications. For example, microjets might not be
easy to adapt to 3D ICs, since it requires the fluid to be injected to the microchannel
from the top.
Compared to single-phase cooling, flow boiling in pin fin-enhanced microgaps
can achieve even higher heat transfer performance. However, new challenges come
from flow instabilities and dryout. Inserting orifice and throttling valve can suppress
the instability. However, they also produce additional pressure drop. The inserted
orifice is usually small and clogging is a significant concern. By depositing or
growing nanostructures on the microchannel surface, flow boiling can be enhanced.
However, the strength of these nanostructures to withstand the flow force at high
flow rates is of concern. Modeling of flow boiling is challenging due to the complex
mechanisms involved. Flow boiling can be divided into different flow regimes with
different heat transfer and bubble dynamics characteristics. Accurate interface heat
and mass transfer models are required.
Before microfluidic cooling can be used widely, cost and reliability issues need to
be considered and addressed. Due to the high efficiency and heat removal capability
of microfluidic cooling, the cooling system can be more compact compared to air
cooling. Liquid-cooling system is usually complicated and requires a flow loop
consisting of pump, heat exchanger, filter, and reservoir. Liquid leakage and clog-
ging need to be addressed. Integrating liquid cooling into an application such as
microprocessor cooling can increase the cost of fabrication, as it would require some
modification of the existing chip fabrication and assembly procedures.
A few recommendations for the future research direction include:

1. Optimization of the liquid-cooled heat sink design to further enhance heat transfer
performance, reduce temperature nonuniformity, and decrease pressure drop
2. Codesign of cooling, power, and signal delivery for 3D ICs. Dynamic control of
power and cooling
3. Exploration of alternative ways to suppress flow boiling instability and enhance
CHF
4. Further integration and miniaturization of the liquid-cooling system
Single- and Multiphase Flow for Electronic Cooling 51

6 Cross-References

▶ Boiling and Two Phase flow in Narrow Channels

▶ Boiling on Enhanced Surfaces
▶ Design of Thermal System
▶ Enhancement of Convective Heat Transfer
▶ Evaporative Heat Exchangers
▶ Forced Convection – Internal Flow
▶ Mixture Boiling
▶ Numerical Methods
▶ Single-Phase Convective Heat Transfer: Fundamental Equations and Founda-
tional Assumptions
▶ Single-Phase Heat Exchangers
▶ Transition and Film Boiling
▶ Two-Phase Heat Exchangers

Nomenclature
A Area [m2]
CHF Critical heat flux [W/m2]
Dp Pin diameter [m]
f Friction factor []
G Maximum mass flux at the smallest cross-sectional area [kg/m2s]
h Heat transfer coefficient [W/m2K]
Hp Pin height [m]
N Number of channels []
Nu Nusselt number []
ΔP Pressure drop [Pa]
R Thermal resistance [K/W]
Sl Liquid source term [kg/m3]
SL Longitudinal spacing [m]
ST Transverse spacing [m]
SV Vapor source term [kg/m3]
t Time [s]
tc Tip clearance [m]
T Temperature [K]
Re Reynolds number []
V Volumetric flow rate [m3/s]
x* Nondimensional streamwise position []

Greek Symbols
α Channel aspect ratio
β Channel width to pitch ratio
ρ Density
52 Y. Joshi and Z. Wan

Subscripts and Superscripts

cond Condensation
cr Cross section
f Fluid
fan Fanning
l Liquid
L Longitudinal
tot Total
sat Saturation
v Vapor

References
Adham AM, Mohd-Ghazali N, Ahmad R (2013) Thermal and hydrodynamic analysis of micro-
channel heat sinks: a review. Renew Sust Energ Rev 21:614–622
Agostini B, Fabbri M, Park JE, Wojtan L, Thome JR, Michel B (2007) State of the art of high heat
flux cooling technologies. Heat Transf Eng 28(4):258–281. doi:10.1080/01457630601117799
Alfieri F, Tiwari MK, Zinovik I, Poulikakos D, Brunschwiler T, Michel B (2010) 3D integrated
water cooling of a composite multilayer stack of chips. J Heat Transf 132(12):121402.
doi:10.1115/1.4002287
Alfieri F, Gianini S, Tiwari MK, Brunschwiler T, Michel B, Poulikakos D (2013) Computational
modeling of hot-spot identification and control in 3-D stacked chips with integrated cooling.
Numer Heat Transf A Appl 65(3):201–215. doi:10.1080/10920277.2013.826480
Ansys [Fluent], release 17.0, Help System (2016), Modeling Multiphase Flows, Ansys, Inc. http://
www.ansys.com/products/academic/citations
Arik M, Bar-Cohen A (2010) Pool boiling of perfluorocarbon mixtures on silicon surfaces. Int J
Heat Mass Transf 53:5596–5604
Bakir MS, Meindl JD (eds) (2008) Integrated interconnect technologies for 3D nanoelectronic
systems. Artech House, Norwood
Bennett DL, Chen JC (1980) Forced convective boiling in vertical tubes for saturated pure
components and binary mixtures. AICHE J 26:454–461. doi:10.1002/aic.690260317
Bogojevic D, Sefiane K, Walton AJ, Lin H, Cummings G, Kenning DBR, Karayiannis TG (2011)
Experimental investigation of non-uniform heating on flow boiling instabilities in a micro-
channels based heat sink. Int J Therm Sci 50(3):309–324
Brunschwiler T, Rothuizen H, Fabbri M, Kloter U, Michel B (2006) Direct liquid jet-impingement
cooling with micron-sized nozzle array and distributed return architecture. In: Proceedings of
ITHERM, pp 196–203
Brunschwiler T, Michel B, Rothuizen H, Kloter U, Wunderle B, Oppermann H, Reichl H (2009)
Interlayer cooling potential in vertically integrated packages. Microsyst Technol 15(1):57–74.
doi:10.1007/s00542-008-0690-4
Chang JY, Kim CU, Michael N, Pathangey B, Gwin P, Prasher R (2006) New electrochemical cell
designs and test methods for corrosion testing of the components in integrated circuit liquid
cooling systems. In: Proceedings of ASME IMECE, pp 257–265
Coskun AK, Ayala JL, Atienza D, Rosing TS (2009) Modeling and dynamic management of 3D
multicore systems with liquid cooling. In: Proceedings of VLSI-SoC, pp 35–40
Dai X, Yang F, Fang R, Yemame T, Khan JA, Li C (2013) Enhanced single- and two-phase transport
phenomena using flow separation in a microgap with copper woven mesh coatings. Appl Therm
Eng 54(1):281–288
Single- and Multiphase Flow for Electronic Cooling 53

Dang B, Bakir MS, Sekar DC, King CR, Meindl JD (2010) Integrated microfluidic cooling and
interconnects for 2D and 3D chips. IEEE Trans Adv Packag 33(1):79–87. doi:10.1109/
TADVP.2009.2035999
Ebadian MA, Lin CX (2011) A review of high-heat-flux heat removal technologies. J Heat Transf
133(11):11080. doi:10.1115/1.4004340
Ekstrand L, Mo Z, Zhang Y, Liu J (2005) Modeling of carbon nanotubes as heat sink fins in
microchannels for microeletronics cooling. In: Proceedings of polytronic, pp 185–187
Ellsworth MJ Jr (2006) Comparing liquid coolants from both a thermal and hydraulic perspective.
Electron Cooling 12(3)
Escher W, Michel B, Poulikakos D (2009) Efficiency of optimized bifurcating tree-like and parallel
microchannel networks in the cooling of electronics. Int J Heat Mass Transf 53(5–6):1421–1430
Escher W, Brunschwiler T, Michel B, Poulikakos D (2010) Experimental investigation of an
ultrathin manifold microchannel heat sink for liquid-cooled chips. J Heat Transf 132
(8):081402. doi:10.1115/1.4001306
Fu Y, Nabiollahi N, Wang T, Wang S, Hu Z, Carlberg B, Zhang Y, Wang X, Liu J (2012) A complete
carbon-nanotube-based on-chip cooling solution with very high heat dissipation capacity.
Nanotechnology 23(4):045304
Fujita Y, Bai Q (1997) Critical heat flux of binary mixtures in pool boiling and its correlation in
terms of Marangoni number. Int J Refrig 20:616–622
Harirchian T, Garimella SV (2012) Flow regime-based modeling of heat transfer and pressure drop
in microchannel flow boiling. Int J Heat Mass Transf 55(4):1246–1260
Harms TM, Kazmierczak MJ, Gerner FM (1999) Developing convective heat transfer in deep
rectangular microchannels. Int J Heat Fluid Flow 20(2):149–157
Hovestreijdt J (1963) The influence of the surface tension difference on the boiling of mixtures.
Chem Eng Sci 18:631–639
Husain A, Kim KY (2008) Shape optimization of micro-channel heat sink for micro-electronic
cooling. IEEE Trans Components Packag Technol 31(2):322–330. doi:10.1109/
TCAPT.2008.916791
Intel Timeline (2016) A history of innovation. http://www.intel.com/content/www/us/en/history/
historic-timeline.html. Accessed 2 Oct 2016
Isaacs S, Kim Y, Mcnamara A, Joshi Y, Zhang Y, Bakir MS (2012) Two phase flow and heat transfer
in pin fin enhanced micro gaps. In: Proceedings of 13th IEEE intersociety conference on thermal
and thermomechanical phenomena in electronics systems (ITherm), pp 1084–1089
Isaacs S, Joshi Y, Zhang Y, Bakir M, Kim Y (2013) Two phase flow and heat transfer in pin fin
enhanced micro gaps with non-uniform heating. In: Proceeding of 4th international conference
on micro/nanoscale heat and mass transfer, pp V001T12A003
Jasperson BA, Jeon Y, Turner KT, Pfefferkorn FE, Qu W (2010) Comparison of micro-pin-fin and
microchannel heat sinks considering thermal-hydraulic performance and manufacturability.
IEEE Trans Components Packag Technol 33(1):148–160. doi:10.1109/TCAPT.2009.2023980
Kandlikar SG (2005) High flux heat removal with microchannels-a roadmap of challenges and
opportunities. Heat Transf Eng 26(8):5–14. doi:10.1115/ICMM2005-75086
Kandlikar SG (2012) History, advances, and challenges in liquid flow and flow boiling heat transfer
in microchannels: a critical review. J Heat Transf 134(3):034001. doi:10.1115/1.4005126
Kandlikar SG (2015) Mechanistic consideration for enhancing flow boiling heat transfer in micro-
channels. J Heat Transf 138(2):021504. doi:10.1115/1.4031648
Kandlikar SG, Alves L (1999) Effects of surface tension and binary diffusion on pool boiling of
dilute solutions: an experimental assessment. J Heat Transf 121:488–493. doi:10.1115/
1.2826008
Kandlikar SG, Bapat AV (2007) Evaluation of jet impingement, spary and microchannel chip
cooling options for high heat flux removal. J Heat Transf Eng 28(11):911–923. doi:10.1080/
01457630701421703
Kandlikar SG, Bulut M (2003) An experimental investigation on flow boiling of ethylene-glycol/
water mixtures. J Heat Transf 125:317–325. doi:10.1115/1.1561816
54 Y. Joshi and Z. Wan

Kandlikar SG, Upadhye HR (2005) Extending the heat flux limit with enhanced microchannels in
direct single phase cooling of computer chips. In: Proceedings of SEMI-THERM, pp 8–15
Kandlikar SG, Widger T, Kalani A, Mejia V (2013) Enhanced flow boiling over open micro-
channels with uniform and tapered gap manifolds. J Heat Transf 135(6):061401. doi:10.1115/
1.4023574
Khan MG, Fartaj A (2010) A review on microchannel heat exchangers and potential applications.
Int J Energy Res 35:553–582. doi:10.1002/er.1720
Khanikar V, Mudawar I, Fisher TS (2009) Flow boiling in a micro-channel coated with carbon
nanotubes. IEEE Trans Components Packag Technol 32(3):639–649. doi:10.1109/
TCAPT.2009.2015232
King CR, Zaveri J, Bakir MS, Meindl JD (2010) Electrical and fluidic C4 interconnections for inter-
layer liquid cooling of 3D ICs. In: Proceedings of ECTC 2010, pp 1674–1681
Kolluri S, Keller S, Denbaars SP, Mishra UK (2011) N-Polar GaN MIS-HEMTs with a 12.1-W/mm
continuous wave output power density at 4 GHz on sapphire substrate. IEEE Electron Device
Lett 32(5):635–637. doi:10.1109/LED.2011.2119462
Konishi CA, Qu W, Pfefferkorn FE (2009) Experimental study of water liquid-vapor two-phase
pressure drop across an array of staggered micropin fins. J Electron Packag 131:021010-
1–021010-8. doi:10.1115/1.3104028
Koo JM, Im S, Jiang L, Goodson KE (2005) Integrated microchannel cooling for three-dimensional
electronic circuit architectures. J Heat Transf 127:49–58. doi:10.1115/1.1839582
Kosar A, Mishra C, Peles Y (2005) Laminar flow across a bank of low aspect ratio micro pin fins. J
Fluid Eng 127:419–430. doi:10.1115/1.1900139
Kosar A, Peles Y (2006a) Thermal-hydraulic performance of MEMS-based pin fin heat sink. J Heat
Transf 128(2):121–131. doi:10.1115/1.2137760
Kosar A, Peles Y (2006b) Convective flow of refrigerant (R-123) across a bank of micro pin fins. Int
J Heat Mass Transf 49:3142–3155
Kosar A, Peles Y (2007) TCPT-2006-096.R2: micro scale pin fin heat sinks-parametric perfor-
mance evaluation study. IEEE Trans Components Packag Technol 30(4):855–865. doi:10.1109/
TCAPT.2007.906334
Kosar A, Schneider B, Peles Y (2011) Hydrodynamic characteristics of crossflow over MEMS-
based pillars. J Fluids Eng 133:081201-1–081201-11. doi:10.1115/1.4004366
Kou HS, Lee JJ, Chen CW (2008) Optimum thermal performance of microchannel heat sink by
adjusting channel width and height. Int Commun Heat Mass Transf 35(5):577–582
Koz M, Ozdemir MR, Kosar A (2011) Parametric study on the effect of end walls on heat transfer
and fluid flow across a micro pin-fin. Int J Therm Sci 50(6):1073–1084
Krishnamurthy S, Peles Y (2008) Flow boiling of water in a circular staggered micro-pin fin heat
sink. Int J Heat Mass Transf 51:1349–1364
Lan J, Xie Y, Zhang D (2011) Flow and heat transfer in microchannels with dimples and pro-
trusions. J Heat Transf 134(2):021901. doi:10.1115/1.4005096
Lee WH (1980) A pressure iteration scheme for two-phase flow modelling. Multiphase Transport
Fundamentals, Reactor Safety Applications, Hemisphere Publishing, Washington, DC
Lee YJ, Lim SK (2008) Co-optimization of signal, power, and thermal distribution networks for 3D
ICs. In: Proceedings of EDAPS, pp 163–166
Lee J, Mudawar I (2009) Low temperature two phase microchannel cooling for high heat flux
thermal management of defense electronics. IEEE Trans Components Packag Technol 32
(2):453–465. doi:10.1109/TCAPT.2008.2005783
Lee W, Son G (2008) Bubble dynamics and heat transfer during nucleate boiling in a microchannel.
Numer Heat Transf A Appl 53(10):1074–1090. doi:10.1080/10407780701789898
Lee PS, Garimella SV, Liu D (2005) Investigation of heat transfer in rectangular microchannels. Int
J Heat Mass Transf 48(9):1688–1704
Li D, Wu GS, Wang W, Wang YD, Liu D, Zhang DC, Chen YF, Peterson GP, Yang R (2012)
Enhancing flow boiling heat transfer in microchannels for thermal management with
Single- and Multiphase Flow for Electronic Cooling 55

monilithically-integrated silicon nanowires. Nano Lett 12(7):3385–3390. doi:10.1021/

nl300049f
Li W, Yang F, Alam T, Peng B, Qu X, Li C (2016) Enhanced flow boiling in microchannels using
auxiliary channels and multiple micronozzles. In: Proceedings of MNHMT 2016, pp
V001T04A010
Lin PH, Fu BR, Pan C (2011) Critical heat flux on flow boiling of methanol – water mixtures in a
diverging microchannel with artificial cavities. Int J Heat Mass Transf 54:3156–3166
Liu M, Liu D, Xu S, Chen Y (2011) Experimental study on liquid flow and heat transfer in micro
square pin fin heat sink. Int J Heat Mass Transf 54:5602–5611
Lorenzini D, Joshi YK (2015) CFD analysis of flow boiling in a silicon microchannel with
non-uniform heat flux. In: Proceedings of ICNMM, pp V001T07A014
Ma A, Wei J, Yuan M, Fang J (2009) Enhanced flow boiling heat transfer of FC-72 on micro-pin-
finned surfaces. Int J Heat Mass Transf 52:2925–2931
Mack CA (2011) Fifty years of Moore’s law. IEEE Trans Semicond Manuf 24(2):202–207.
doi:10.1109/TSM.2010.2096437
Madhour Y, d’Entremont BP, Marcinichen JB, Michel B, Thome JR (2014) Modeling of two phase
evaporative heat transfer in three dimensional multicavity high performance microprocessor
chip stacks. J Electron Packag 136(2):021006. doi:10.1115/1.4027436
Magnini M, Thome JR (2015) Use of two-phase CFD simulations to develop a boiling heat transfer
prediction method for slug flow within microchannels. In: Proceedings of IPACK, pp
V003T10A020
McGillis WR, Carey VP (1996) On the role of Marangoni effects on the critical heat flux for pool
boiling of binary mixtures. J Heat Transf 118:103–109. doi:10.1115/1.2824021
Mcneil DA, Raeisi AH, Kew PA, Bobbili PR (2010) A comparison of flow boiling heat transfer in
in-line mini pin fin and plane channel flows. Appl Therm Eng 30(16):2412–2425
Mita JR, Qu W, Kobayashi MH, Pfefferkorn FE (2011) Experimental study and numerical analysis
of water single-phase pressure drop across an array of circular micro-pin-fins. In: Proceedings of
ASME/JSME 2011 8th thermal engineering joint conference, pp T10131–T10131-9
Mohapatra SC (2006) An overview of liquid coolants for electronics cooling. Electron Cooling 12
(5):1–6
Moore GE (1998) Cramming more components onto integrated circuits. In: Proceedings of the
IEEE, pp 82–85
Moores KA, Joshi Y (2003) Effect of tip clearance on the thermal and hydrodynamic performance
of a shrouded pin fin array. J Heat Transf 125:999–1006. doi:10.1115/1.1621897
Moores K, Kim J, Joshi Y (2009) Heat transfer and fluid flow in shrouded pin fin arrays with and
without tip clearance. Int J Heat Mass Transf 52(25–26):5978–5989
Morshed AKMM, Yang F, Ali MY, Khan JA, Li C (2012) Enhanced flow boiling in a microchannel
with integration of nanowires. Appl Therm Eng 32:68–75
Mudawar I (2011) Two phase microchannel heat sinks: theory, applications, and limitations. J
Electron Packag 133:041002. doi:10.1115/1.4005300
Muwanga R, Hassan I, MacDonald R (2007) Characteristics of flow boiling oscillations in silicon
microchannel heat sinks. J Heat Transf 129(10):1341–1351. doi:10.1115/1.2754946
Naphon P, Khonseur O (2008) Study on the convective heat transfer and pressure drop in the
microchannel heat sink. Int Commun Heat Mass Transf 36(1):39–44
Ochende TB, Liebenberg L, Meyer JP (2007) Constructual cooling channels for micro-channel heat
sinks. Int J Heat Mass Transf 50(21–22):4141–4150
Pan Z, Weibel JA, Garimella SV (2015) A cost-effective modeling approach for simulating phase
change and flow boiling in microchannels. In: Proceedings of IPeACK, pp V003T10A023
Park JE, Thome JR, Michel B (2009) Effect of Inlet orifice on saturated CHF and flow visualization
in multi-microchannel heat sinks. In: Proceedings of SEMI-THERM, pp 1–8
Peles Y, Kosar A, Mishra C, Kuo CJ, Schneider B (2005) Forced convective heat transfer across a
pin fin micro heat sink. Int J Heat Mass Transf 48:3615–3627
56 Y. Joshi and Z. Wan

Peng XF, Peterson GP, Wang BX (1996) Flow boiling of binary mixtures in microchannel plates. Int
J Heat Mass Transf 39:1257–1264
Prasher R (2006) Thermal interface materials: historical perspective, status, and future directions.
In: Proceedings of the IEEE, pp 1571–1586
Prasher RS, Dirner J, Chang JY, Myers A, Chau D, He D, Prstic S (2007) Nusselt number and
friction factor of staggered arrays of low aspect ratio micropin-fins under cross flow for water as
fluid. J Heat Transf 129:141–153. doi:10.1115/1.2402179
Qu W, Mudawar I (2004) Transport phenomena in two-phase micro-channel heat sinks. J Electron
Packag 126(2):213–224. doi:10.1115/1.1756145
Qu W, Siu-Ho A (2009) Experimental study of saturated flow boiling heat transfer in an array of
staggered micro-pin-fins. Int J Heat Mass Transf 52:1853–1863
Qu W, Mudawar I, Lee SY, Wereley ST (2006) Experimental and computational investigation of
flow development and pressure drop in a rectangular micro-channel. J Electron Packag 128:1–9.
doi:10.1115/1.2159002
Rahman MM (2000) Measurements of heat transfer in microchannel heat sinks. Int Commun Heat
Mass Transf 27(4):495–506
Reeser A, Bar-Cohen A, Hetsroni G (2014) High vapor quality two phase heat transfer in staggered
and inline micro pin fin arrays. In: Proceedings of 14th IEEE intersociety conference on thermal
and thermomechanical phenomena in electronics systems (ITherm), pp 213–222
Robinson AJ (2009) A thermal-hydraulic comparison of liquid microchannel and impinging liquid
jet array heat sinks for high-power electronics cooling. IEEE Trans Components Packag Technol
32(2):347–357. doi:10.1109/TCAPT.2008.2010408
Rozati A, Tafti DK, Blackwell NE (2008) Effect of pin tip clearance on flow and heat transfer at low
Reynolds numbers. J Heat Transf 130:071704. doi:10.1115/1.2909184
Saenen T, Thome JR (2015) Novel dynamic numerical microchannel evaporator model to investi-
gate parallel channel instabilities. In: Proceedings of IPACK 2015, pp V003T10A021
Salamin YI, Hu SX, Hatsagortsyan KZ, Keitel CH (2006) Relativistic high-power laser-matter
interactions. Phys Rep 427(2–3):41–155
Samalam VK (1989) Convective heat transfer in microchannels. J Electron Mater 18(5):611–617.
doi:10.1007/BF02657475
Sathyanarayana A (2013) Pool and flow boiling of novel heat transfer fluids from nanostructured
surfaces. PhD dissertation, Georgia Institute of Technology, Atlanta, GA, USA 30332
Sekar D, King C, Dang B, Spencer T, Thacker H, Joseph P, Bakir M, Meindl J (2008) A 3D-IC
technology with integrated microchannel cooling. In: Proceedings of IITC, pp 13–15
Shafeie H, Abouali O, Jafarpur K (2010) Numerical investigation of heat transfer enhancement in a
microchannel with offset micro pin fins. In: Proceedings of FEDSM-ICNMM, pp 1–6
Shi B, Srivastava A, Wang P (2011) Non-uniform micro-channel design for stacked 3D-ICs. In:
Proceedings of ACM/EDAC, pp 658–663
Short BE Jr, Raad PE, Price DC (2002a) Performance of pin fin cast aluminum coldwalls, part II:
colburn j factor correlations. J Therm-Phys Heat Transf 16(3):397–403. doi:10.2514/2.6693
Short BE Jr, Raad PE, Price DC (2002b) Performance of pin fin cast aluminum coldwalls, part I:
friction factor correlations. J Therm-Phys Heat Transf 16(3):389–396. doi:10.2514/2.6692
Siu-Ho A, Qu W, Pfefferkorn F (2007) Experimental study of pressure drop and heat transfer in a
single-phase micropin-fin heat sink. J Electron Packag 129:479–487. doi:10.1115/1.2804099
Sparrow EM, Ramsey JW, Altemani CA (1980) Experiment on in-line pin fin arrays and perfor-
mance comparisons with staggered arrays. J Heat Transf 102:44–50. doi:10.1115/1.3244247
Sridhar A, Vincenzi A, Ruggiero M, Brunschwiler T, Atienza D (2010) 3D-ICE: fast compact
transient thermal modeling for 3D ICs with inter-tier liquid cooling. In: Proceedings of ICCAD,
pp 463–470
Steinke ME, Kandlikar SG (2004) Review of single-phase heat transfer enhancement techniques for
application in microchannels, minichannels and microdevices. Int J HeatTechnol 22(2):3–11
Steinke ME, Kandlikar SG (2006) Single phase liquid friction factors in microchannels. Int J Therm
Sci 45:1073–1083
Single- and Multiphase Flow for Electronic Cooling 57

Sui Y, Teo CJ, Lee PS, Chew YT, Shu C (2010) Fluid flow and heat transfer in wavy microchannels.
Int J Heat Mass Transf 53(13–14):2760–2772
Sung MK, Mudawar I (2008) Effects of jet pattern on single-phase cooling performance of hybrid
micro-channel/micro-circular-jet-impingement thermal management scheme. Int J Heat Mass
Transf 51(19–20):4614–4627
Sung MK, Mudawar I (2009) CHF determination for high-heat flux phase change cooling system
incorporating both micro-channel flow and jet impingement. Int J Heat Mass Transf 52
(3–4):610–619
Tan SP, Toh KC, Khan N, Pinjala D, Kripesh V (2011) Development of single phase liquid cooling
solution for 3-D silicon modules. IEEE Trans Components Packag Manuf Technol 1
(4):536–544. doi:10.1109/TCPMT.2010.2100710
Tang GY, Tan SP, Khan N, Pinjala D, Lau JH, Yu AB, Vaidyanathan K, Toh KC (2010) Integrated
liquid cooling systems for 3D stacked TSV modules. IEEE Trans Components Packag Technol
33(1):184–195. doi:10.1109/TCAPT.2009.2033039
Tuckerman DB, Pease RFW (1981) High performance heat sinking for VLSI. IEEE Electron
Device Lett 2(5):126–129. doi:10.1109/EDL.1981.25367
Tullius JF, Tullius TK, Bayazitoglu Y (2012) Optimization of short micro pin fins in minichannels.
Int J Heat Mass Transf 55:3921–3932
Van Wijk WR, Vos AS, Van Strallen SJD (1956) Heat transfer to boiling binary liquid mixtures.
Chem Eng Sci 5:68–80
Wan Z, Xiao H, Joshi Y, Yalamanchili S (2014a) Co-design of multicore architectures and
microfluidic cooling for 3D stacked ICs. Microelectron J 45(2):1814–1821
Wan Z, Yueh W, Joshi Y, Mukhopadhyay S (2014b) Enhancement in CMOS chip performance
through microfluidic cooling. In: Proceedings of 2014 20th international workshop on thermal
investigations of ICs and systems (THERMINIC), pp 1–5
Wang XQ, Mujumdar AS, Yap C (2007) Effect of bifurcation angle in tree-shaped microchannel
networks. J Appl Phys 102:073530. doi:10.1063/1.2794379
Wang ZH, Wang XD, Yan WM, Duan YY, Lee DJ, Xu JL (2011) Multi-parameters optimization for
microchannel heat sink using inverse problem method. Int J Heat Mass Transf 54
(13–14):2811–2819
Warrier GR, Kim CJ, Ju YS (2014) Microchannel cooling device with perforated side walls: design
and modeling. Int J Heat Mass Transf 68:174–183
Webb RL (2005) Next generation devices for electronic cooling with heat rejection to air. J Heat
Transf 127:2–10. doi:10.1115/1.1800512
Wei XJ, Joshi YK, Ligrani PM (2006) Numerical simulation of laminar flow and heat transfer inside
a microchannel with one dimpled surface. J Electron Packag 129(1):63–70. doi:10.1115/
1.2429711
Wu Z, Sunden B (2014) On further enhancement of single-phase and flow boiling heat transfer in
micro/minichannels. Renew Sust Energ Rev 40:11–27
Xie G, Chen Z, Sunden B, Zhang W (2013a) Numerical analysis of flow and thermal performance
of liquid-cooling microchannel heat sinks with bifurcation. Numer Heat Transf Part A: Appl 61
(11):902–919. doi:10.1080/10407782.2013.807689
Xie G, Liu Y, Sunden B, Zhang W (2013b) Computational study and optimization of laminar heat
transfer and pressure loss of double-layer microchannels for chip liquid cooling. J Therm Sci
Eng Appl 5:011004. doi:10.1115/1.4007778
Yang F, Li X, Li W, Li C (2015) Integrate monolithic nanostructures in microchannels to enhance
flow boiling heat transfer of HFE-7000. In: Proceedings of IPACK, pp V002T06A006
Zhang HY, Pinjala D, Joshi YK, Wong TN, Toh KC (2003) Development of liquid cooling
techniques for flip chip ball grid array packages with high heat flux dissipations. IEEE Trans
Components Packag Technol 28:127–135. doi:10.1109/TCAPT.2004.843164
Zhang Y, King CR, Zaveri J, Kim YJ, Sahu V, Joshi Y, Bakir MS (2011a) Coupled electrical and
thermal 3D IC centric microfluidic heat sink design and technology. In: Proceedings of ECTC,
pp 2037–2044
58 Y. Joshi and Z. Wan

Zhang T, Wen JT, Peles Y, Catano J, Zhou R, Jense MK (2011b) Two- phase refrigerant flow
instability analysis and active control in transient electronics cooling systems. Int J Multiphase
Flow 37(1):84–97
Zhang H, Liu JJ, Li Y, Yao SC (2013) Porous media modeling of two-phase microchannel cooling
of electronic chips with nonuniform power distribution. In: Proceedings of IMECE, pp
V08BT09A071
Zhu Y, Antao DS, Chu KH, Chen S, Hendricks TJ, Zhang T, Wang EN (2016) Surface structure
enhanced microchannel flow boiling. J Heat Transf 138(9):091501. doi:10.1115/1.4033497