International Journal of Heat and Mass Transfer 184 (2022) 122300

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: [Link]/locate/hmt

Achieving ultra-high coefficient of performance of two-phase

microchannel heat sink with uniform void fraction
Xingchi Jiang a, Shiwei Zhang b, Yuanjie Li a, Zuankai Wang a, Chin Pan a,∗
a
Department of Mechanical Engineering, City University of Hong Kong, Kowloon, Hong Kong, China
b
Intelligent Manufacturing Engineering Laboratory of Functional Structure and Device in Guangdong, School of Mechanical and Automotive Engineering,
South China University of Technology, Guangzhou 510640, China

a r t i c l e i n f o a b s t r a c t

Article history: Thermal management of high-power electronics/devices has been a very challenging issue for data cen-
Received 22 April 2021 ters, flourishing electric vehicle revolution as well as cooling of the first wall of a Tokamak fusion reactor.
Revised 16 October 2021
As a prominent cooling strategy, two-phase flow in microchannels has received extensive research over
Accepted 21 November 2021
decades. However, inherent defects such as relatively low critical heat flux due to the dry out of liquid
Available online 4 December 2021
film near the exit of the channel and relatively high two-phase flow pressure drop at high heat flux still
Keywords: hinder extensive commercial applications. The present work proposes a robust microchannel heat sink
Thermal management design toward uniform void fraction distribution along the flow direction. The two-phase flow pressure
Counter flow diverging microchannels drop will not increase significantly with increase in heat flux. This is made possible by an innovative
Flow boiling and unique combination of diverging microchannel and counter-flow manifold, which enables extensive
Uniform void fraction channel-to-channel heat transfer, especially near the end or inlet of the channels. Moreover, flow visu-
Coefficient of performance
alization reveals that nearly uniform void fraction distribution along the channel with relatively high
bubble slug ratio is possible. For a relatively large area of 12 cm2 and under the limitation of wall tem-
perature of 140 °C, a heat flux as high as 3525 kW/m2 is achieved without sign of reaching the critical
heat flux with nearly negligible pressure drop increment comparing to single-phase convection for a rel-
atively low mass flux of 600 kg/m2 s. An ultra-high coefficient of performance of 75,675 is attained under
a low inlet temperature and low mass flux.
© 2021 Elsevier Ltd. All rights reserved.

1. Introduction operation. Besides, subjected to Moore’s law [4], the miniaturiza-

tion of electronics, i.e., integrated circuit chips [5], and super-fast
Facing the challenges of energy consumption and management, electric power conversion devices, i.e. IGBT, photovoltaic converter
technological innovation has always been the driving force of [6–8] call for intricate thermal management problems. High perfor-
mankind’s never-ending progress [1]. Demand and contradiction mance and compactness of thermal management system for high-
arise along with each other in the modern society. For example, power electronics are of critical concern for manufacturers. In fore-
the fast-growing supercomputing centers, the flourishing IT indus- seeable future, consuming more energy is inevitable for the ther-
try, and the thriving electric vehicle revolution have greatly im- mal management of ultra-high power electronics; therefore, the ef-
proved the work efficiency and life quality of human, however, ficient and reliable cooling strategies are ultimate goals to pursue
more stringent and complex requirements are brought for energy by researchers.
management, storage and conversion. As a large amount of wa- Two-phase heat transfer with its potentially high heat transfer
ter and electricity has been consumed for the cooling systems of capability is widely expected to be a new generation [9] of high-
electronics and devices [2], increasing attention has been paid on performance heat dissipation approach to meet the rapid devel-
the development of ultra-effective thermal management systems opment of the electronics industry. Flow boiling in microchannels
[3]. Currently, it becomes crucial for the semi-conductor devices [10–13], as one of the promising modes in thermal management,
to pursue high-performance while maintaining low energy cost on has been extensively examined both experimentally [14–17] and
numerically [18–21] for several decades. However, with inherent
defects caused by violent phase change in confined space, this
∗
Corresponding author. cooling technology is still hindered from commercial applications.
E-mail address: chinpan@[Link] (C. Pan). In general, annual flow with thin liquid film prevails in microchan-

[Link]
0017-9310/© 2021 Elsevier Ltd. All rights reserved.
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

parameters have important effects on two-phase flow heat trans-

Nomenclature fer performance [29,30], the control of flow pattern is limited and
passive by only setting operation parameters in conventional mi-
Ab base area of heat sink, m2 crochannels. It is very hard to manipulate the bubble behavior in
Ac inlet cross-section area of the microchannels, m2 microchannels.
CHF critical heat flux, kW/m2 Most flow boiling experiments are conducted in a small-scale
COP coefficient of performance base substrate, i.e. 1 cm2 , for analysis and evaluation; however,
Cp specific heat, J/kgK many high-power electronic devices are with larger area, i.e.
d copper substrate thickness, m 10 cm2 –30 cm2 [7,31,32]. The availability is worthy to be further
Dh hydraulic diameter, m investigated, as the longer length of microchannel provides more
G mass flux, kg/m2 s continuous heating power if assuming the heat flux is uniformly
HTC/ ht p two-phase heat transfer coefficient, kW/m2 K applied along channel length. Longer microchannels may lead to
kcu thermal conductivity of copper, W/mK worse thin-film evaporation process in downstream without timely
Lb total bubble length in visualization area, m liquid replenishment. Thus, the risk of hot-spot/partial dry-out will
Lv microchannel length in visualization area, m rise.
m˙ mass flow rate, kg/s Numerous efforts have been dedicated to address such short-
ONB onset of nucleate boiling comings of convective boiling in microchannels. Currently, the
Pin pressure at inlet, kPa mainstream strategies include surface enhancement, upstream
Pout pressure at outlet, kPa throttling, downstream expanding, vapor venting, auxiliary jetting
Qe f f heat transfer rate to fluid, W nozzles, etc. By enhancing the surface morphology, micro/nano-
Qloss heat loss rate, W scale coatings [29,33–35] may provide more active nucleation sites,
Q pump pumping power, W promote the bubble nucleation earlier and departure away from
Qtotal total heat input rate to heat sink, W the heating surface more frequently. However, there still exists a
qb heat flux imposed on base area, kW/m2 contingent risk of nanoparticles delamination in long-term run-
Re Reynolds number ning [36,37], which may cause deterioration of heat transfer char-
Ta ambient temperature, °C acteristics and accompanying problems of escaping particles into
Tin inlet temperature, °C the cooling liquid. For example, Choi et al. [37] conducted sub-
Tout outlet temperature, °C cooled flow boiling using deionized water in microchannels de-
Ts heat sink temperature, °C posited with Fe3 O4 nanoparticles. Part of the coating detached
Tsat saturation temperature, °C from the surface base during the violent flow boiling accompa-
Tw wall surface temperature, °C nied with drastic oscillations of both pressure and temperature, es-
v coolant velocity, m/s pecially under the conditions of critical heat flux (CHF) and large
V˙ volumetric flow rate, m3 /s mass fluxes. A similar phenomenon was also reported by Morshed
et al. [38]. The nano-coating was not robust enough to sustain
Greek symbols
harsh condition during their experiments, and lots of nanoparti-
ρ coolant density, kg/m3
cles detached from wall surface. Moreover, the micro/nano-coating
μ coolant dynamic viscosity, Pa•s
on the channel surface may result in higher flow resistance and
 difference of parameters
therefore pumping power penalty. For example, Morshed et al.
φ bubble slug ratio
[39] conducted flow boiling tests in a rectangular microchannel
β wall temperature non-uniformity
coated with copper nanowires using FC-72 as working medium.
Although much better heat dissipation performance of CHF and
nels at high heat fluxes due to the rise in void fraction. The par- heat transfer coefficient (HTC) were attained, this strategy failed
tial dry out of the liquid film triggers the critical heat flux, even to manage flow boiling instability fundamentally and ineffective to
though the vapor quality at the channel exit is still well below lower the pressure drop.
unity corresponding to the complete evaporation of liquid in the Inlet restriction is widely utilized for two-phase flow instabil-
channel. Moreover, the increase of void fraction with heat flux also ity suppression [40–42]. Mukherjee and Kandlikar [40] proved that
causes steep increase in two-phase flow pressure drop due to fric- placing restrictions at channel inlet can significantly prevent explo-
tion as well as fluid acceleration. The relatively low critical heat sive bubble growth and alleviate flow reversal in microchannels.
flux and high two-phase flow pressure drop limit the applications Similarly, Wang et al. [41] conducted flow boiling tests in parallel
of convective boiling in microchannels. Moreover, violent volume microchannels with a length of 30 mm and a hydraulic diameter of
expansion usually occurs nearly instant; therefore, the rapid oscil- 186 um. Their results indicated the microchannels without inlet re-
lations of temperature and pressure in confined channel passage strictions experienced temperature and pressure oscillations when
lead to the flow instability [22–24], especially under extremely bubble rapidly expanded, whereas steady flow boiling without os-
high heat flux conditions. The two-phase flow instability usually cillations was achieved in microchannels with triangular throttling
causes back flow and premature dry out, which lead towards the restrictor. However, inlet throttling is commonly accompanied with
failure or even destruction of the cooling system [25–28]. significantly extra pressure penalty.
As is well-known, the two-phase flow provides more uniform Expanding/diverging microchannels [43] perform a merit on
temperature distribution than single-phase cooling due to the stabilizing flow state without extra pressure drop increment. Pan
smaller temperature variation in the saturation temperature and and his co-workers [44–48] first conducted a series of two-phase
heat transfer coefficient during nucleate boiling and forced con- flow experiments in diverging microchannels. The diverging de-
vective evaporation. However, further improvement on wall tem- sign along the flow direction provides extra space for liquid-vapor
perature variation is still subject to operation parameters (i.e., inlet transition, and it is effective to mitigate the bubble slug back-
subcooling, mass flux). The inlet section is usually with lower wall ward expansion. Later, Balasubramanian et al. [49] examined the
temperature than the downstream wall surface, and such a non- advantages of diverging microchannels in their contrast studies.
uniform trend could be enlarged if a higher inlet subcooling is set Compared with straight microchannels, the expanding microchan-
or longer microchannel length is adopted. Although the operation nels attained ∼30% lower pressure drop and lower wall temper-

2
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

ature with more significant flow stability. Similar results were phase flow pressure drop is suppressed. Moreover, the two-phase
also reported in their later work of stepping fined microchannels flow will be much more stable and the temperature distribution
[50], which provided extra vertical space for vapor volume expan- will be more uniform.
sion. However, one drawback of diverging microchannels in width Indeed, our preliminary results [59] shows that, comparing with
is that the non-uniform channel wall thickness may lead to in- the traditional co-current flow design, the CFDM achieved ∼45.1%
effective use of base area. increment in two-phase HTC, ∼73.8% reduction of pressure drop
By separating coolant vapor flow from the confined space, va- and ∼123.1% increment of coefficient of performance (COP) on the
por venting [51–53] achieves desirable improvement to stabilize same substrate area. The present study looks for an optimal opera-
the two-phase flow and lower the pressure drop penalty in mi- tion condition for a more uniform two-phase flow structure along
crochannels. For example, David et al. [51] reported flow boiling the channel through more studies with lower inlet temperature.
experiments in microchannels with a 220 nm pore, hydrophobic Extensive investigations (i.e., operation parameters, bubble behav-
PTFE membrane. They achieved nearly 60% lower of pressure drop iors and heat transfer performance) are conducted for better un-
and 4.4 °C reduction in the average substrate temperature un- derstanding the counter flow diverging microchannels, including
der the same condition, which was attributed to that the venting flow patterns, HTC, pressure penalty, COP, and temperature unifor-
membrane device can vent the vapor phase into separate transport mity under various test conditions. This novel strategy has unique
channels and relieve pressure surge. Recently, Mohiuddin et al. features of two-phase heat transfer performance and great poten-
[53] investigated the effects of vapor venting in microchannels tial for thermal management of high-power electronics/devices.
with different aspect ratios, their results reported that ∼30% incre-
ment of the two-phase HTC accompanied with a 10% pressure drop 2. Experimental setup
reduction was attained by the lowest aspect ratio channel with va-
por venting membrane. Despite of the merits on pressure drop re- Fig. 1 illustrates a schematic diagram of the experimental loop,
duction and flow stabilizing effect, the additional vapor transport which includes a gear pump (FG300, Fluid-o-Tech), a control valve,
structure leads to more complex structure and higher fabrication a filter (5 um), a flowmeter (OM004, FLOMEC), a water-bath pre-
cost. heater, a test section, a condenser, and a reservoir tank. Outside
These above approaches of enhancing the heat transfer perfor- the loop, a chiller provides cooling water for continuous operation
mance of microchannel heat sink may not address the early trigger of the condenser. The pump provides driving force for circulating
of critical heat flux due to the partial dry out of liquid film. flow, while the control valve adjusts the flow rate. The flowme-
Since microchannel manifold and structure have significant im- ter measures the volumetric flow rate of the working fluid, while
pacts on heat transfer performance [43], multiple inlet injection is the front filter removes contaminants and protects the flowmeter
proposed as an innovative strategy with segmented inlet coolant and test section. The water-bath preheater provides working fluid
replenishment for microchannels [54–57]. For example, Radwan with a specific inlet temperature. After the test section, the con-
et al. [55] conducted two-phase cooling tests using ethanol and denser collects the expelled liquid-vapor mixture, which will be
Novec-70 0 0 on a monolithic double-layer microchannel heat sink chilled by cooling water. Then, the condensed coolant flows back
for cooling a solar cell. Their results exhibited the microchannels to the reservoir, where the working medium is stored. Thus, a
with bidirectional inlet arrangement can reduce the maximum cell complete test loop is formed. There is a venting valve upon the
temperature, maintain more uniform temperature distribution, and reservoir for keeping the inner pressure under normal ambient
further enhance the electrical efficiency of the solar cell. Li et al. condition. Besides, a T-type thermocouple is inserted to monitor
[58] illustrated a hybrid strategy of microchannels with integrat- the coolant temperature there. To measure the heat dissipation
ing multiple microscale nozzles connected to auxiliary channels tests, a synchronized data acquisition system (DAQ) is built, in-
on sidewalls. Attributed to continuous liquid supply through four cluding: a data acquisition collector (PXIle-1082, National Instru-
evenly distributed micronozzles, they achieved a high CHF and ments), a self-designed LabVIEW program and a personal com-
the dry out phenomenon in microchannels was delayed. Indeed, puter. Temperature and pressure drop are recorded by T-type ther-
the multiple inlets installed along the microchannel, especially in mocouples and differential pressure transducer, respectively. Ad-
downstream, can provide extra replenishment for harsh condition ditionally, the typical dynamic two-phase flow patterns are cap-
of thin film evaporation. However, additional coolant passage ar- tured by a high-speed camera (Crash Cam-1540, IDT) for further
rangement may occupy a part of effective heat transfer area for analysis.
two-phase transition. Fig. 2(a) illustrates the counter-flow diverging microchannel
The present study intends to remedy the defects of microchan- (CFDM) heat sink with the dimension of 40 mm × 30 mm
nel heat sink through a robust design. Thus, the heat transfer per- (length × width). The SEM image for the channel bottom surface
formance and critical heat flux can be significantly enhanced and illustrates machined scratches by CNC machining process. There
yet the two-phase flow pressure drop increment with heat flux are no or very few cavities and particles presented. By adopting
can be minimized by making the void fraction distribution nearly vertical inlet design, 25 microchannels in total are arranged on the
uniform in the microchannels via a counter flow diverging mi- copper base plate, 13 in one direction and 12 in the counter direc-
crochannel (CFDM) heat sink with three-dimensional inlet design. tion. For each microchannel, the length is 39.7 mm, the inlet width
Through the channel-to-channel heat transfer via counter flow is 0.5 mm and the depth is 3 mm. Besides, a diverging angle of
subcooled boiling in diverging microchannel heat sink, the void 0.5° is adopted for each channel while the channel wall thickness
fraction and, therefore, two-phase flow structure can be manipu- remains constant of 0.3 mm. Thus, the channel outlet width devel-
lated. A nearly uniform void fraction and two-phase flow pattern ops to 1.2 mm. The main reason for this manifold is that, diverg-
in the microchannel heat sink are achievable. Bubble nucleation in ing angle in single channel provides extra space for the coolant
the highly subcooled upstream region is possible through inten- volume expansion during the phase change process making two-
sive heat transfer from the downstream portion of both neighbor- phase flow stable. Meanwhile, such a counter-flow array brings sig-
ing channels. On the other hand, the heat transferring to the en- nificant merits of active heat exchange between neighboring chan-
trance region of both neighboring channels reduces or suppresses nels making uniform void fraction in each channel possible. Such a
the growth of void fraction near the exit. Consequently, the critical unique and innovative combination will significantly enhance heat
heat flux may be delayed near the channel exit, heat transfer co- transfer coefficient and critical heat flux, while minimize the pres-
efficient is increased through the optimized flow pattern and two- sure drop increment with heat flux.

3
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

Fig. 1. Schematic diagram of test loop.

Fig. 3(b) shows the distribution of 9 T-type thermocouples, strate of microchannels is made of pure copper for its excellent
which are inserted into the copper heat sink at 3.5 mm below the thermal conductivity and sturdiness. Above the copper heat sink,
channel bottom surface, on the heat sink. Specifically, Ts6 , Ts7 , Ts8 , borosilicate glass is customized as the cover glass for visualization
Ts9 , reflect the general temperature distribution of the heat sink, while sealing the sink top. A stainless-steel frame is installed on
while Ts2 , Ts4 , can verify the symmetric distribution corresponding the top of test module to fix all components by bolts and nuts.
to Ts7 , Ts8 , respectively. Besides, Ts1 , Ts3 , Ts5 , are located at the cen- Below the test section, the heating base, as shown in Fig. 6(e),
tral line to illustrate the distribution of central temperature along simulates a real high-power electronic device with high heat dissi-
the flow direction. pation demand and well matches with the target heat sink. A thin
The top view and side view of the CFDM sample with dimen- layer of liquid gallium-based alloy with high thermal conductiv-
sions are illustrated in Fig. 2(c). The arrows show the inlet coolant ity is adopted between the heat sink and heating base to reduce
flow direction. Such a coolant flow organization contributes to the the contact thermal resistance. The heating source is provided by
countercurrent flow in the microchannels. The heated two-phase 6 inserted cartridge heaters (1 Kw × 6), and the heating power is
mixture at the outlet will be collected at both ends. controlled by a linear regulator and monitored by a DC power me-
The housing module, as shown in Fig. 4(d), includes several ter. Fiberglass is fully filled in the gap between the stainless-steel
main components: housing (stainless steel), thermal isolation plate shell and the copper heating block in order to minimize heat loss.
(polyphenylene sulfite, PPS), liquid distribution plate (polypheny-
lene sulfite, PPS) and two cover-glass layers (borosilicate glass). The 3. Procedures and data reduction
housing body provides reliable structural support for the whole
test section, meanwhile, two chambers are integrated on the hous- 3.1. Experiment procedures
ing for the inlet coolant and the outlet heated vapor-liquid mix-
ture collection. For each chamber, two reserved holes are set on To perform the tests, the test module is assembled upon the
the chamber sides, one for the inserted T-type thermocouple, and heating block and connected to the flow loop. The de-ionized wa-
the other for pressure measuring probe. The thermal isolation plate ter is adopted as the working medium. The inlet temperature and
and liquid distribution plate are fabricated to lead the coolant the flowrate are set by the water-bath preheater and the gear
flowing in designated direction, as shown succinctly in Fig. 5(e). pump, respectively. Before each test, rigorous degassing of water
The working fluid liquid can be evenly distributed into each mi- is conducted by more than one-hour of boiling in the reservoir.
crochannel through vertical inlet holes upon two edges of the heat The temperatures and pressures are recorded by the DAQ system
sink, after flowing and being heated along the channel, the hori- after the test section achieves a steady state, which is character-
zontal outlet plenum around the copper sink will collect the two- ized by a negligible change in heat sink central temperature, i.e.,
phase mixture and expel it to the outlet chamber. With high ther- Ts3 , is less than 0.5 °C over a 3 min period. Since the present tests
mal resistance and good mechanical performance, the PPS ensures are conducted in a wide range of operation conditions, a consis-
the sealing and robustness during high temperature tests. The sub- tent limit of maximum wall temperature for safety concern is set

4
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

Fig. 2. (a) counter-flow diverging microchannel heat sink and Scanning Electron Microscope (SEM) image of bare copper surface; (b) distribution of T-type thermocouples
on the heat sink; (c) top view plane, local dimension diagram and side view section of CFDM sample; (d) CFDM heat sink with housing components; (e) heating base and
schematic diagram of flow direction.

as Tw- max ≤ 140 °C for each experiment. High-speed visualization Table 1

Operating parameters for the present experiments.
of two-phase flow pattern is synchronously recorded with temper-
ature and pressure measurements. In the present study, the sam- Operating parameters
pling frequency of temperature and pressure is 1 Hz for 1 min, and Mass flux (kg/m2 s) 200, 400, 600
the high-speed flow visualization is conducted at 40 0 0 frames per Inlet temperature (°C) 50, 60, 70, 80, 90
second. The present study on CFDM heat sink is investigated with Effective heat flux (kW/m2 ) 60–3525
a wide range of operating parameters of inlet temperature, mass
flux at inlet and effective heat flux on the base area of the heat
sink, as summarized in Table 1. The pressure at the channel exit power can be accurately measured. Thus, the effective base heat
is kept at 1 atm, while the maximum wall temperature is kept at flux is then given by:
about 140 °C for the safety of the test section.
Qe f f
qb = (5)
Ab
3.2. Data reduction
where Ab is the base area of the microchannel substrate, Ab = W ×
The present study adopts the same methodology as our prelim- L, in which W and L are the width and length of the copper heat
inary study [59]. The heat loss and therefore the effective heating sink, respectively.

5
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

Fig. 3. Bubble evolvement in a single microchannel while heat flux is increasing under condition of Tin = 80 °C, G = 200 kg/m2 s corresponding to different heat flux. (a)
boiling curve; (b) qb ” = 274 kW/m2 , bubble grows after nucleation and shrinks while flowing to near the exit; (c) qb ” = 539 kW/m2 , bubble grows to become slug flow; (d)
qb ” = 1054 kW/m2 , fully-developed slug flow; (e) qb ” = 1678 kW/m2 , Tw ≈140 °C, slug flow with slightly back flow.

The local wall temperature corresponding to the measured local is defined as the average of the inlet and outlet absolute pressures:
heat sink temperature can be calculated by one-dimensional heat
conduction assumption as: Pave = (Pin + Pout )/2 (8)
q d
Tw,i = Ts,i − b , i = 1, 2, 3...9 (6) Notably, the pressure drop may rise as heating power increases;
kcu
consequently, the saturation temperature corresponding to the sys-
where d denotes the copper substrate thickness between the tem pressure may also varies with heating power.
thermocouple location and bottom surface of the microchannels, The high-speed camera records the bubble behaviors within the
kcu denotes the copper thermal conductivity, and Ts,i denotes the lo- visualization window. To accommodate the unique vertical inlet
cal heat sink temperature, directly measured by the corresponding design in the present study, the CFDM heat sink is covered with
T-type thermocouple. the PPS distribution plate above, the available length range for vi-
The two-phase heat transfer coefficient can then be calculated sualization is about 28 mm, which is only 70% of the full chan-
as: nel length of 40 mm. For a particular image, the bubble slug in
q b each channel can be identified using a commercial software, Im-
ht p = (7)
Tw − Tsat ageJ, based on the color threshold difference [60,61]. Consequently,
the length of the bubble slug in each channel can be measured
where Tw is the maximum wall temperature, which is correspond-
and analyzed. A statistics chart exhibiting the slug bubble distribu-
ing to the middle section temperature of the heat sink. Notably,
tion under various test conditions may be obtained. In particular, a
unlike the conventional single-direction microchannels, which usu-
bubble slug ratio is defined as:
ally presents unidirectional increasing temperature distribution
along channel length direction, the middle section temperature of φ = Lb /Lv N (9)
the CFDM is higher than that of inlet and outlet section, owing to
the counter-flow coolant arrangement. As maintaining the maxi- where Lb denotes the summation of bubble slug length in each
mum temperature is a priority issue for protecting the target elec- channel identified and calculated by ImageJ, Lv is the microchan-
tronic device, the present study adopts the highest wall tempera- nel length in the visualization window, and N presents the quan-
ture for boiling curve and heat transfer coefficient. tity of microchannels (N = 25). The bubble slug length is highly
The pressure drop P across the microchannels is directly mea- related to the void fraction and two-phase flow heat transfer in a
sured by the differential pressure transducer. The system pressure microchannel heat sink.

6
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

Fig. 4. Bubble slug ratio as a function of heat flux and inlet temperature under mass flux of (a) G = 200 kg/m2 s, (b) G = 400 kg/m2 s, and (c) G = 600 kg/m2 s under
the limitation of maximum wall temperature Tw ≈140 °C. Nearly uniform void fraction is demonstrated for bubble slug ratio over 60%. (Note: inlet flow and outlet flow are
marked as blue arrow and red dash arrow, respectively). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this
article.)

7
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

Fig. 5. Boiling curves with wall superheat based on the highest wall temperature Tw3 for different inlet temperatures for (a) G = 200 kg/m2 s, (b) G = 400 kg/m2 s, (c)
G = 600 kg/m2 s; (d) boiling curve for different mass fluxes/Reynolds numbers for inlet temperature of 60 °C; (e) boiling curve in fully developed boiling region independent
of mass flux and inlet temperature; (f) two-phase HTC versus base heat flux for G = 600 kg/m2 s.

8
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

Fig. 6. Pressure drop versus base heat flux under various mass fluxes of (a) G = 200 kg/m2 s, (b) G = 400 kg/m2 s, (c) G = 600 kg/m2 s; (d) coefficient of performance (COP)
under various conditions with Tw- max ≈140 °C.

The performance of a forced convective heat sink may be quan- dard for discussion:
tified by comparing the heat dissipating rate with the pump- ρvDh
ing power. Following the similar evaluation in previous literature Re = (13)
μ
[2,62,63], the coefficient of performance in the cooling system may
be defined as: where ρ is coolant density, v is coolant velocity at inlet, Dh is inlet
hydraulic diameter of a single microchannel, μ is viscosity for a
COP = Qe f f /Q pump (10) certain inlet temperature.

where Qpump is the pumping power to drive the working fluid

through the CFDM heat sink with a volume flow rate V˙ : 3.3. Uncertainty analysis

Q pump = PV˙ (11) Measurement uncertainties mainly include heating power, tem-
perature, pressure drop, and flow rate. For the heating power, the
The temperature non-uniformity may be quantitatively mea- digital DC power meter yields an uncertainty of ± 1.0%. Measured
sured as the ratio of the largest wall temperature difference to the by T-type thermocouple, the temperature error is ± 0.4% (in range
maximum wall temperature: of 125–350°C); the pressure drop uncertainty of the transducer is
Tw−max − Tw−min ± 0.1%. The volume flow rate measured by volumetric flow meter
β= (12) is ± 1.0%. The uncertainty due to the mechanical machining fabri-
Tw−max
cation for the present heat sink is ± 0.02 mm, which is provided
where Tw- max (°C) and Tw- min (°C) denote the maximum and mini- by manufacturer. As ImageJ software is applied for bubble slug
mum wall temperature on the heat sink surface. identification and length statistics, the measurement error comes
As diverging angle is adopted in present CFDM heat sink, the from pixel color threshold set when the software identifies bubble
inlet Reynolds number is selected and calculated as a unified stan- slug boundary, the uncertainty of one single bubble slug length is

9
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

Table 2 If the heat flux imposed is further increased to a very high heat
Uncertainties of major parameters.
flux of qb ”= 1678 kW/m2 , Fig. 3(e) depicts the evolvement of two-
Main Parameters Uncertainties phase flow patterns. Bubbles nucleate much faster near the chan-
Mass flux G ± 4.18% nel inlet, and the slug bubble is elongated further. Significantly, the
Base heat flux qb ” ± 5.84% bubble begins to expand to upstream periodically resulting in mi-
Wall superheat Tw -Tsat ± 5.88% nor back flow. Such a phenomenon illustrates that the vigorous
Heat transfer coefficient htp ± 8.28% evaporation exhibits greater heat dissipation performance. Mean-
Bubble slug ratio φ ± 2.12%
while, this leads to the thinning of liquid film at microchannel
Pressure drop P ± 0.10%
Coefficient of performance ± 5.92% downstream. Therefore, under such a high heat flux, the imposed
heat rate from the microchannel bottom is larger than the channel-
to-channel heat transfer effect at the channel exit under consider-
ation.
about ± 1.5%. Refer to the method proposed by Kline and McClin-
For traditional co-current flow microchannels, it commonly
tock [64], the uncertainty propagations of the main parameters can
shows a premature CHF when the void fraction is still relatively
be calculated, as listed in Table 2 below:
low. This is because the continuous heating along the microchan-
nel will result in void expansion, and possibly partial dry out due
4. Results and discussion to the lack of liquid replenishment at downstream section. How-
ever, by adopting counter flow microchannel design, the lateral
4.1. Flow patterns heat transfer through the channel walls can provide several benefi-
cial effects to alleviate such an inherent problem in microchannels.
The present counter flow diverging microchannel array demon- Firstly, the subcooled liquid in the CFDM entering through the
strates unique development of two-phase flow pattern through inlet of both neighboring channels can not only remove all or part
heating power imposed from the base area and channel-to-channel of heat from the bottom surface near the channel exit, but also
heat transfer from the side walls. Fig. 3 illustrates the evolution of reduce the growth of the void fraction and limit the rapid bub-
bubbles in a microchannel of the CFDM heat sink for some heat ble expansion. If the heat flux imposed is low, the lateral heat
fluxes along the boiling curve shown in Fig. 3(a) corresponding transfer rate could be larger than that imposed from the bottom
to Tin = 80 °C and G = 200 kg/m2 s. Notably, the CFDM exhibits surface and there could be vapor condensation instead resulting
unique bubble behavior of growing after nucleation and shrinking in the shrinkage of bubbles. Even under the high heat flux con-
while flowing toward the exit, as shown in Fig. 3(b). Under a rela- ditions, such a unique heat transfer mechanism through the mi-
tive low heat flux of qb ”= 274 kW/m2 , two tiny bubbles are nucle- crochannel wall can still play a significant role in restraining the
ated at the middle stream at some time which is set as t = 0 ms. bubble expansion at the downstream region; therefore, the mi-
These two bubbles grow and merge with each other to form an crochannel downstream section can keep a stable thin liquid film
elongated bubble. This bubble grows further to reach its maximum without partial dry out. Consequently, the critical heat flux may be
volume at t = 50 ms. After that, its volume is rapidly shrinking in enhanced.
the following 40 ms while it is approaching the exit. Before flow- On the other hand, for the inlet region, such lateral heat trans-
ing out of the visualization window, this typical bubble slug has fer from the neighboring channels will supplement the heating
already disappeared. This unique phenomenon demonstrates the effect imposed on the bottom wall and promote the rapid tem-
strong lateral chilling effect from the neighboring microchannels. perature elevation of inlet subcooled coolant and make the bub-
Near the exit of the channel, in addition to the heat imposed ver- ble nucleation at a more upstream location. From a macroscopic
tically on the bottom wall from the heating block, there is strong view, the heat exchange between outlets and adjacent inlets on
lateral heat transfer to the adjacent two channels with subcooled the CFDM array can contribute to more uniform temperature dis-
liquid flow from the vertical hole to the narrow inlet of the chan- tribution on the whole heating surface. Highly stable flow states
nel. The shrinkage of bubble near the exit in Fig. 3(b) with a low are observed in all the groups under conditions of G = 400 and
heat flux indicates that such lateral heat transfer rate to the neigh- 600 kg/m2 s, even for a very high heat flux of 3525 kW/m2 . (A
bor channels is, in fact, larger than the heat flow rate imposed sample video can be viewed in Supplementary “Fully developed
from the bottom wall. boiling in counter flow diverging microchannels (G = 400 kgm-2s,
With the increasing heat flux, the bubbles nucleate at a more Tin = 60 °C, qb = 2797 kWm-2)”).
upstream location, as shown in Fig. 3(c) for a heat flux of As illustrated earlier, the bubble slug ratio is the fraction of to-
539 kW/m2 . The tiny bubbles generate and depart from the wall tal slug length to the total channel length in the visualization win-
surface, then grow faster due to a higher heat flux imposed at dow. The slug ratio is somewhat larger than the void fraction, as
channel bottom. After that, these bubbles merge with each other there is liquid film between the slug bubble and channel. The bub-
to form slug bubble in the confined microchannel. The slug bub- ble slug ratio reflects the vapor volume generated through the net
ble maintains a short length until it flows out of the visualization heat transfer from the bottom and side walls. Moreover, the length
window smoothly. This suggests that the heat flow rate imposed at and state of the vapor slug are directly related to the heat transfer
the channel bottom is close to the lateral heat transfer rate to the characteristics and the two-phase flow in microchannel. When the
adjacent channels. bubble slug ratio is small, the two-phase flow resistance is small.
Fig. 3(d) illustrates the bubble evolvement when a larger heat The void fraction and the boiling heat transfer intensity may also
flux of qb ”=1054 kW/m2 is imposed. The nucleation sites move fur- be low if there are only a small number of bubbles. On the other
ther upstream to a location closer to the channel inlet. The bubbles hand, the thin liquid film between the wall and a long bubble slug
develop quickly and coalesce with downstream vapor slug more inside has high evaporative heat transfer intensity. The divergent
rapidly. During this fully developed boiling process, the slug bub- design and the coupled channel-to-channel heat transfer near the
ble expands to be longer than that under low heat flux condition. channel exit may reduce evaporation or even condense the vapor
Notably, the bubble slug length remains approximately unchanged inside the bubble and enable the slug bubble to flow smoothly
during this visualization period. This suggests that the lateral chill- through the channel.
ing effect is competitive with vertical heat flux imposed near the Based on the bubble evolution illustrated in Fig. 3, further
exit of the channel under examination. statistics of bubble slug has been acquired to illustrate a general

10
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

map of the two-phase flow development. Fig. 4 illustrates the bub- wall superheat due to more thermal energy is needed for pre-
ble slug ratio in the CFDM as a function of heat flux and inlet tem- heating cooler inlet liquid and activating bubble nucleation. This
perature for the three flow rates in the present study. For the flow is consistent with that in co-current microchannels [29,30]. In the
pattern statistics, the lengths of bubble slugs are well identified two-phase region, the boiling curves with different inlet subcool-
and calculated by ImageJ software with necessary manual inspec- ings gradually tend to be closer with each other, this suggests the
tion, in case of wrongly identification of bubble boundary as the appearance of fully developed boiling (FDB). The fully developed
shadows inevitably exist in a few microchannel corners. For each subcooled boiling is coincident with the bubble slug ratio being
test, 5 photos during 1 s are randomly selected, and 660 images in greater than about 40%.
total are processed to exhibit the general trend of bubble slug ra- Fig. 5(d) investigates the influence of mass flux and Reynolds
tio. Generally, the figures indicate that for a specific heat flux the number on heat transfer characteristics for inlet temperature of
bubble slug expands in confined microchannels when the heat flux 60 °C. In single-phase regime, the larger Reynolds number ob-
or inlet temperature is increased. It is understood that more heat- viously brings benefit on heat convection performance, the heat
ing power imposed on the two-phase flow and less lateral heat flux increases almost in proportion to Reynolds number, mean-
transfer effect from the neighboring channels in the downstream while, the three curves maintain steady spacing from each other.
section will enhance bubble growth rate and result in longer bub- Once the ONB occurs, these curves with different Reynolds num-
ble slug length. bers gradually narrow the spacing, which means when the boil-
Four different flow pattern images corresponding to their bub- ing heat transfer mechanism is involved, the influence of Reynolds
ble slug ratios are illustrated near Fig. 4(a). For example, when number will become weakened during the initial boiling stage. Fur-
bubble slug ratio is 13.0% for Tin = 50 °C, qb ” = 1136 kW/m2 , the ther increasing heat flux, the three curves converge towards a sim-
small bubble slugs are randomly distributed in the CFDM heat sink, ilar slope in the fully developed boiling regime, the increase of
the two-phase flow is very stable with little impediment effect by Reynolds number can no longer bring the proportional increment
the vapor slug. As for the bubble slug ratio with medium values of of heat transfer capacity. Here, the growth of liquid film evapora-
49.5% and 58.3%, these two flow patterns exhibit elongated bub- tion rate in two-phase flow mainly governs the heat transfer per-
bles flowing smoothly in microchannels. Meanwhile, the liquid thin formance of microchannels.
film evaporation may dominate the heat transfer mechanism dur- Fig. 5(e) illustrates the heat transfer performance in the FDB re-
ing this stage. The novel countercurrent coolant flow design with gion, in which bubble slug flow prevails consistently and the HTCs
neighboring chilling effect contributes to such a uniform void frac- are relatively independent of mass flux and inlet temperature, as
tion formed in microchannels. If extremely high heat flux and/or shown in Fig. 5(d) and Fig. 5(f), respectively. The boiling curves
high inlet temperature are set, which will significantly weaken the under various operation conditions tend to gather together. This
neighboring chilling effect, the bubble slug ratio may increase to a is similar to fully developed subcooled boiling in ordinarily sized
high level. Here, the flow pattern image of ϕ = 84.3% for Tin = 90 °C channels. The heat flux increases exponentially with the wall su-
is with minor oscillation. perheat and can be fitted well by the following equation:
Overall, Fig. 4(a) for G = 200 kg/m2 s shows that more data  
qb = 1.47(Tsat )2.35 kW/m2 (14)
points in the dark red zone. There are only two cases experience
minor back flow with the bubble ratios of 83.4% and 84.6%, respec- Heat transfer coefficient is a critical issue for thermal man-
tively. The range under 80% can be treated as highly stable zone agement, especially for the cooling of microelectronics. Fig. 5(f)
with desirable smooth flow and high heat transfer performance in presents the effect of inlet temperature on HTC for the case with
the CFDM design. For the above two cases with very high bub- the mass flux of 600 kg/m2 s. The figure indicates that HTC in-
ble slug ratio, i.e., void fraction, there is still no sign of dry out. creases with decrease in inlet temperature and demonstrates the
This demonstrates a significant merit of such counter flow design: highest at the ONB. During ONB, nucleate boiling is characterized
homogenizing the void fraction in the channel and avoid the oc- with frequent departure of tiny bubbles and a large amount of
currence of dry out near the channel exit as in traditional design heat is taken away through enhanced convection due to bubble
of co-current flow. Therefore, the present design will significantly agitation and latent heat of evaporation with a small wall super-
enhance the critical heat flux. This may revolutionize the design of heat resulting in an HTC as high as 334 kW/m2 K for the case with
microchannel heat sink or heat exchanger. an inlet temperature of 50 °C. The sudden decrease in HTC af-
Fig. 4(b) and (c) exhibits that under the mass flux of ter ONB is due to limited active nucleation sites and the increas-
G = 400 kg/m2 s and 600 kg/m2 s, all bubble slug ratios are be- ing heat flux after ONB will make the small bubbles rapidly grow
low 70% and 60%, respectively. In particular, for the test conditions and merge together to form slug bubble with lower heat trans-
with the lowest inlet temperature of Tin =50 °C under the highest fer capability possibly due to less bubble agitation effects. Such a
mass flux of G = 600 kg/m2 s in the present study, the bubble slug steep reduction in heat transfer coefficient after ONB have also re-
ratios range from 1.3% to17.1% only. This suggest that there is still ported in many previous studies [29,58,65–67]. Moreover, the fig-
a great potential for the CFDM to reach much higher high flux if ure clearly indicates a high inlet subcooling enhances two-phase
there is no limitation of Tw ≈140 °C. HTCs. For example, for G = 600 kg/m2 s and maintaining the same
limitation of Tw- max =140 °C, the peak value of HTC corresponding
4.2. Heat transfer performance to ONB for Tin = 90 °C reaches 92.7 kW/m2 K only, while it can
reach 334 kW/m2 K for Tin = 50 °C. The nearly 260% amplification
Fig. 5 illustrates the heat transfer performance with bubble slug of two-phase HTC is brought by decreasing the inlet temperature,
ratio indicated for various inlet temperatures and mass fluxes in which extends the small bubble region and induces more channel-
the present study. As shown in Fig. 5(a–c), in the single-phase re- to-channel heat interaction. The HTC begins to rise with increase
gion the base heat flux increases slowly with increasing wall super- in heat flux after fully developed subcooled boiling is established
heat and significantly with decreasing the inlet temperature. These with the stable slug flow, possibly due to the thinning of liquid
curves surge after the onset of nucleate boiling (ONB) with no- film between the slug bubble and heating wall and more intensive
table steeper slopes, which may be attributed to the heat transfer evaporation.
performance enhancement brought by phase change with large la- The stronger channel-to-channel heat transfer for a lower in-
tent heat dissipation and bubble agitation effects. As the inlet tem- let temperature enables a significantly higher maximum heat flux
perature decreases, the ONB is gradually moving toward a higher for a given wall temperature limit of 140 °C in the present study.

11
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

For G = 200 kg/m2 s and Tin = 50 °C, the maximum heat flux With such a small bubble slug ratio, one may expect randomly
can reach 2286 kW/m2 , which is 77.9% higher than that for Tin distributed short slug bubbles in the heat sink. Moreover, such
=90 °C of qb ” = 1285 kW/m2 . The similar trend prevails for the short slug bubbles may be shrinking near the channel exit due
cases of G = 400 kg/m2 s and G = 600 kg/m2 s. Fig. 5(b) indicates to the chilling effect from neighboring channels. Therefore, it is of
the elevation of maximum heat flux is 79.7%, from 1773 kW/m2 no surprise that the pressure rise for this case is almost negligi-
for Tin = 90 °C to 3186 kW/m2 for Tin = 50 °C. Besides, Fig. 5(c) ble. For the highest inlet temperature of 90 °C, Fig. 4(c) indicates
shows less increment of 69.3% from 2083 kW/m2 to 3525 kW/m2 , that the largest bubble slug ratio is about 46% under the limita-
because the two-phase flow boiling is still under developing with tion of Tw ≈140 °C. The flow pattern images near Fig. 4(a) illustrate
the bubble slug ratio smaller than 40% according to Fig. 4(c). This smooth slug flow for such range of bubble slug ratio. With the high
desirable enhancement of heat dissipation capacity brought by dif- mass flow rate in the diverging channel, one may expect a thicker
ferent inlet subcooling in the CFDM significantly exceeds that of liquid film between the bubble and channel wall, especially near
conventional microchannels [30,31]. the channel exit with chilling effects from the neighboring chan-
nels. Therefore, the slug bubble may flow through the channel very
4.3. Pressure drop and COP smoothly with insignificant two-phase pressure drop.
This is a remarkable achievement for a convective boiling sys-
Usually the large two-phase pressure drop increment in mi- tem. The cooling system with convective boiling is commonly suf-
crochannels with increase in heat flux is caused by the rapid void fered from the severe pressure drop surge after ONB, especially
fraction expansion, and the vapor slug continuously aggravates the for microchannel systems, which leads to significant energy con-
impediment effect in confined microchannels as heat flux increases sumption for pumping the working fluid. Moreover, this may also
[23,27]. Fig. 6 illustrates the pressure drop of the CFDM under var- cause two-phase flow instability. The counter-flow diverging mani-
ious experimental conditions. Under a relatively low mass flux of fold provides an innovative, passive and yet robust way to address
G = 200 kg/m2 s, Fig. 6(a) indicates that, in general, a lower in- this problem.
let temperature shows slightly higher pressure drop during initial Fig. 6(d) illustrates the coefficient of performance, which is de-
single-phase convection, but lower two-phase pressure drop during fined as the heat dissipation rate to the corresponding pumping
two-phase flow. For each inlet temperature, the single-phase pres- power (see Eq. (10)), as a function of mass flux and inlet tem-
sure drop decreases slightly with increase in heat flux before the perature under the condition of the maximum possible heat flux
ONB. Moreover, the higher the inlet temperature, the smaller the corresponding to the wall temperature limit of 140 °C. The figure
pressure drop. This is because of the reduction of liquid viscosity indicates that for a given mass flux, the COP is approximately lin-
as temperature rises, which leads to the decrease in frictional re- early proportional to the inlet subcooling. On the other hand, for
sistance. a given inlet subcooling, the COP increases with decrease in mass
After ONB, notable increments of two-phase pressure drop for fluxes. A large mass flux results in a lower COP, as the pump-
most cases, especially for the case of Tin = 90 °C, 80 °C and 70 °C ing power is approximately proportional to the third power of the
and eventually the two-phase pressure drop for Tin = 90 °C be- flowrate. Consequently, the case of G = 200 kg/m2 s presents the
comes the highest for a given heat flux among all inlet tempera- highest COP. When Tin = 90 °C, the COP is 34,189, which is far
tures. The increasing trend is similar to that for convective boiling more beyond that from the single-phase cooling strategy [2,63]. As
in the conventional microchannels, but the increment with heat the inlet temperature decreasing to Tin = 50 °C, the COP demon-
flux is significantly smaller, especially for the case with the lowest strates an extremely high value of 75,675 for a corresponding inlet
inlet temperature of 50 °C, for which the pressure drop changes subcooling of about 65.5 °C. This further shows another superior
from 3.82 kPa at the lowest heat flux with single phase convection merit brought by the CFDM design among two-phase flow cooling
to 4.78 kPa with boiling two-phase flow. This is due to the strong strategy [62].
channel-to-channel heat transfer at such a low inlet temperature. As demonstrated in flow patterns in Section 4.1, the present de-
The bubble slug ratio is decreased, or void fraction growth is sup- sign provides unique lateral channel-to-channel heat transfer ef-
pressed as shown earlier in Fig. 4. fect from the hot two-phase mixture near the exit of the chan-
The above general trend of pressure drop also prevails in nel to the subcooled liquid near the inlet of the neighboring chan-
the groups of G = 400 kg/m2 s and 600 kg/m2 s (in Fig. 6(b) nels. Fig. 7(a) and (b) illustrates the possible heat transfer paths
and Fig. 6(c), respectively). The most cases experience a minor near one end of the heat sink. The vertical heat flux is uniformly
reduction trend with heat flux in the single-convection region. imposed on the sink bottom from the heating module. There is
The cooler inlet liquid fed-in performs a slightly higher pressure also significant heat transfer in the horizontal direction from the
penalty. However, different from the cases for G = 200 kg/m2 s, the hot fluid near the channel exit to the subcooled fluid in the en-
pressure drops for these cases with a higher mass flux exhibit only trance region. Moreover, such vertical flow may also produce sig-
a small increment even for high heat fluxes. Generally, the pres- nificant impingent cooling effect to the bottom surface at the inlet.
sure drops through the present CFDM can be quite independent If the inlet temperature is low enough and/or the mass flux is large
of heat flux. The pressure drop in the two-phase flow regime is enough, such an impinging cooling effect together with the lat-
very close to that in the single-phase regime. In fact, the pressure eral heat transfer rate could be larger than the heat transfer rate
drop increments are ranging in 4.7%−13.2% for G = 400 kg/m2 s, from the bottom surface. Notably, the channel is wider near the
and 2.7%−9.6% for G = 600 kg/m2 s, respectively. These excellent exit due to its diverging nature. Such thermal interaction prevails
results demonstrate that the common pressure drop surge in the intensively on two edges of heat sink, as depicted in Fig. 7(b), and
two-phase regime has been well-suppressed in the present design gradually diminishes along the axial direction, finally it reaches a
of counter-flow diverging microchannels. This must be attributed temperature equilibrium state in the middle section. Such channel-
to the neighboring chilling effect making the void fraction distri- to-channel heat transfer will promote the void generation in the
bution uniform in the channel. entrance region while suppress the void growth in the exit region.
Using Fig. 6(c) with G = 600 kg/m2 s as an example, the in- Consequently, a uniform void fraction distribution along the chan-
significant pressure rise after ONB may be understood with the nel is possible, as shown in Fig. 7(c). Such uniform void fraction
help of bubble slug ratio diagram shown in Fig. 4(c). For the low- will prevent the dry out and so occurrence of the CHF near the
est inlet temperature of 50 °C, the largest bubble slug ratio cor- channel exit while leaving a thick liquid film upstream. Moreover,
responding the highest heat flux of 3525 kW/m2 is about 17%. the common increment of two-phase flow pressure drop with heat

12
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

Fig. 7. Schematic illustration of channel-to-channel heat transfer effect on the counter flow diverging microchannel heat sink. (a) 3D inlet flow design provides counter flow
as well as impinging cooling effect, (b) channel-to-channel heat transfer effect because the temperature difference between the near entrance of a channel and near exit of
both neighboring channels, (c) more uniform void fraction distribution along each channel under fully-develop condition may be resulted.

Fig. 8. Wall temperature distribution maps under different inlet temperatures of (a) Tin = 50 °C, (b) Tin = 70 °C, (c) Tin = 90 °C for G = 600 kg/m2 s and Tw ≈140 °C, (d)
temperature non-uniformity under various conditions.

flux will be totally suppressed or minimum. Thus, the CFDM ar- 4.4. Temperature uniformity
ray employs the large temperature difference between neighboring
channels creating superior merits to solve inherent flaws in mi- Maintaining the temperature uniformity of microelectronic de-
crochannels: relatively low critical heat flux and very high two- vices is crucial for the thermal management to prevent signif-
phase flow pressure drop. icant temperature mal-distribution. As is well-known, the con-

13
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

vective boiling may make the wall temperature more uniform. CRediT authorship contribution statement
However, subcooled flow boiling in co-current flow microchannels
is inherently limited with cool entrance and hot outlet region. Xingchi Jiang: Investigation, Data curation, Writing – origi-
The channel-to-channel heat transfer through the CFDM mani- nal draft. Shiwei Zhang: Investigation. Yuanjie Li: Investigation.
fold design may effectively reduce temperature non-uniformity, as Zuankai Wang: Supervision, Project administration. Chin Pan:
demonstrated in our preliminary work through the measurement Conceptualization, Supervision, Project administration, Writing –
of wall temperatures at three locations along the central axial di- original draft.
rection [59]. The two-dimensional distribution of thermocouples in
the present study enables a further investigation on temperature Acknowledgment
uniformity.
Fig. 8(a–c) illustrate the channel wall temperature distribution. The present study was sponsored by internal projects 9380091
Based on the measurements and linear fitting using software, Ori- and 7005233 of the City University of Hong Kong (CityU).
gin, the 2-D wall temperature distribution maps can be attained
for G = 600 kg/m2 s and inlet temperature of 50 °C, 70 °C and Supplementary materials
90 °C, respectively. Note that each map is constructed for the cases
with the maximum wall temperature of about 140 °C. These fig- Supplementary material associated with this article can be
ures clearly demonstrate warm center and cool peripheral. High found, in the online version, at doi:10.1016/[Link].
inlet subcooling and high flow rate will exaggerate such situation. 2021.122300.
The channel-to-channel heat transfer has significantly reduced the
temperature non-uniformity considering highly subcooled fluids at Reference
the inlet.
Fig. 8(d) illustrates the temperature non-uniformity as a [1] W. Kuo, C. Pan, A reliability look at energy development, Joule 2 (1) (2018)
5–9.
function of inlet subcooling and mass flux. For example, for
[2] R. van Erp, R. Soleimanzadeh, L. Nela, G. Kampitsis, E. Matioli, Co-designing
G = 600 kg/m2 s, the maximum temperature difference is only electronics with microfluidics for more sustainable cooling, Nature 585 (7824)
about 10 °C when Tin = 90 °C, then it gradually increases as inlet (2020) 211–216 Sep.
[3] N. Jones, How to stop data centres from gobbling up the world’s electricity,
temperature decreases and finally reaches to nearly 20 °C under
Nature 561 (2018) 163–166.
Tin = 50 °C. Consequently, the temperature non-uniformity ranges [4] S. Krishnan, S.V. Garimella, G.M. Chrysler, R.V. Mahajan, Towards a thermal
from 8% to 14%, as shown in Fig. 8(d). The similar trend appears Moore’s law, IEEE Trans. Adv. Packag. 30 (3) (2007) 462–474.
for G = 200 kg/m2 s with temperature non-uniformity ranging from [5] S.V. Garimella, et al., Thermal challenges in next-generation electronic systems,
IEEE Trans. Compon. Packag. Technol. 31 (4) (2008) 801–815.
7% to 11% and G = 400 kg/m2 s from 5% to 7%, respectively. Signif- [6] I. Mudawar, D. Bharathan, K. Kelly, S. Narumanchi, Two-phase spray cooling of
icantly, the CFDM design limits the temperature non-uniformity to hybrid vehicle electronics, IEEE Trans. Compon. Packag. Technol. 32 (2) (2009)
below 15% even for the largest mass flux of 600 kg/m2 s and low- 501–512.
[7] P. Wang, P. McCluskey, A. Bar-Cohen, Two-phase liquid cooling for thermal
est inlet temperature 50 °C is set. This demonstrates another sig- management of IGBT power electronic module, J. Electron. Packag. 135 (2)
nificant merit of CFDM heat sink in mitigating wall temperature (2013).
mal-distribution. [8] N. Gilmore, V. Timchenko, C. Menictas, Microchannel cooling of concentrator
photovoltaics: a review, Renew. Sustain. Energy Rev. 90 (2018) 1041–1059 Jul.
[9] I. Mudawar, Recent advances in high-flux, two-phase thermal management, J.
5. Conclusions Therm. Sci. Eng. Appl. 5 (2) (2013).
[10] A. Radwan, S. Ookawara, S. Mori, M. Ahmed, Uniform cooling for concen-
trator photovoltaic cells and electronic chips by forced convective boiling in
The present study demonstrates a high-performance heat sink 3D-printed monolithic double-layer microchannel heat sink, Energy Convers.
with uniform void fraction distribution along each channel achiev- Manag. 166 (2018) 356–371.
able combining counter-flow design with diverging microchannels [11] K.L. Zhang, Z.H. Liu, B.C. Zheng, A new 3D chip cooling technology using mi-
cro-channels thermosyphon with super-moist fluids and nanofluids, Energy
through channel-to-channel heat transfer effects. Large mass flux
Convers. Manag. 128 (2016) 44–56.
and low inlet temperature provide significant lateral heat trans- [12] Z. Azizi, A. Alamdari, M.R. Malayeri, Convective heat transfer of Cu–water
fer from hot fluid near the channel exit to the cool inlet fluid nanofluid in a cylindrical microchannel heat sink, Energy Convers. Manag. 101
(2015) 515–524.
in the two neighboring channels. This provides a robust way for
[13] Z. Wu, B. Sundén, On further enhancement of single-phase and flow boiling
more uniform void fraction distribution and stable two-phase flow heat transfer in micro/minichannels, Renew. Sustain. Energy Rev. 40 (2014)
in the microchannels. Such a microchannel heat sink with near 11–27.
uniform void fraction enables a very high heat transfer capabil- [14] W. Ling, W. Zhou, W. Yu, F. Zhou, J. Chen, K.S. Hui, Experimental investiga-
tion on thermal and hydraulic performance of microchannels with interlaced
ity with remarkably low pressure drop. Under the wall temper- configuration, Energy Convers. Manag. 174 (2018) 439–452.
ature limit of about 140 °C in the present study, a heat flux of [15] C. Baldassari, M. Marengo, Flow boiling in microchannels and microgravity,
3525 kW/m2 is achieved in the CFDM with a relatively large area Prog. Energy Combust. Sci. 39 (1) (2013) 1–36.
[16] S. Bigham, A. Fazeli, S. Moghaddam, Physics of microstructures enhancement
of 12 cm2 . Significantly, there is no sign of reaching the dry out of thin film evaporation heat transfer in microchannels flow boiling, Sci. Rep.
situation indicating that there is large room toward the critical 7 (2017) 44745 Mar 17.
heat flux. Owing to adjacent wall chilling effect near the two ends [17] M. Mehrvand, S.A. Putnam, Transient and local two-phase heat transport at
macro-scales to nano-scales, Commun. Phys. 1 (1) (2018).
of the heat sink, the stable flow can be maintained with a very [18] M. Dehghan, M.S. Valipour, S. Saedodin, Microchannels enhanced by porous
low two-phase pressure drop. The pressure increments with heat materials: heat transfer enhancement or pressure drop increment? Energy
flux from single-phase convection to two-phase regime are quite Convers. Manag. 110 (2016) 22–32.
[19] C. Leng, X.D. Wang, T.H. Wang, W.M. Yan, Optimization of thermal resistance
small under G = 200 kg/m2 s, and even nearly negligible under
and bottom wall temperature uniformity for double-layered microchannel heat
G = 400 and 600 kg/m2 s. An ultra-high coefficient of performance sink, Energy Convers. Manag. 93 (2015) 141–150.
of 75,675 is achieved. Such a revolutionized heat sink is scalable, [20] Z. Shi, T. Dong, Entropy generation and optimization of laminar convective
heat transfer and fluid flow in a microchannel with staggered arrays of pin
easy-fabricated, robust, and potentially useful for thermal manage-
fin structure with tip clearance, Energy Convers. Manag. 94 (2015) 493–504.
ment of high-power electronics/devices. [21] V. Leela Vinodhan, K.S. Rajan, Computational analysis of new microchannel
heat sink configurations, Energy Convers. Manag. 86 (2014) 595–604.
Declaration of Competing Interest [22] S. Kakac, B. Bon, A review of two-phase flow dynamic instabilities in tube boil-
ing systems, Int. J. Heat Mass Transf. 51 (3–4) (2008) 399–433.
[23] Y.K. Prajapati, P. Bhandari, Flow boiling instabilities in microchannels and their
Authors declare that there is no conflict of interest. promising solutions – a review, Exp. Therm. Fluid Sci. 88 (2017) 576–593.

14
X. Jiang, S. Zhang, Y. Li et al. International Journal of Heat and Mass Transfer 184 (2022) 122300

[24] H. Grzybowski, R. Mosdorf, Dynamics of pressure drop oscillations during flow [46] P.H. Lin, B.R. Fu, C. Pan, Critical heat flux on flow boiling of methanol–water
boiling inside minichannel, Int. Commun. Heat Mass Transf. 95 (2018) 25–32. mixtures in a diverging microchannel with artificial cavities, Int. J. Heat Mass
[25] H.Y. Wu, P. Cheng, H. Wang, Pressure drop and flow boiling instabilities Transf. 54 (15–16) (2011) 3156–3166.
in silicon microchannel heat sinks, J. Micromech. Microeng. 16 (10) (2006) [47] B.R. Fu, M.S. Tsou, C. Pan, Boiling heat transfer and critical heat flux of
2138–2146. ethanol–water mixtures flowing through a diverging microchannel with arti-
[26] K.H. Chang, C. Pan, Two-phase flow instability for boiling in a microchannel ficial cavities, Int. J. Heat Mass Transf. 55 (5–6) (2012) 1807–1814.
heat sink, Int. J. Heat Mass Transf. 50 (11–12) (2007) 2078–2088. [48] B.R. Fu, C.Y. Lee, C. Pan, The effect of aspect ratio on flow boiling heat transfer
[27] W. Qu, I. Mudawar, Measurement and prediction of pressure drop in of HFE-7100 in a microchannel heat sink, Int. J. Heat Mass Transf. 58 (1–2)
two-phase micro-channel heat sinks, Int. J. Heat Mass Transf. 46 (15) (2003) (2013) 53–61.
2737–2753. [49] K. Balasubramanian, P.S. Lee, L.W. Jin, S.K. Chou, C.J. Teo, S. Gao, Experimen-
[28] C.J. Kuo, Y. Peles, Pressure effects on flow boiling instabilities in parallel mi- tal investigations of flow boiling heat transfer and pressure drop in straight
crochannels, Int. J. Heat Mass Transf. 52 (1–2) (2009) 271–280. and expanding microchannels – a comparative study, Int. J. Therm. Sci. 50 (12)
[29] D. Deng, W. Wan, H. Shao, Y. Tang, J. Feng, J. Zeng, Effects of operation parame- (2011) 2413–2421.
ters on flow boiling characteristics of heat sink cooling systems with reentrant [50] K. Balasubramanian, P.S. Lee, C.J. Teo, S.K. Chou, Flow boiling heat transfer and
porous microchannels, Energy Convers. Manag. 96 (2015) 340–351. pressure drop in stepped fin microchannels, Int. J. Heat Mass Transf. 67 (2013)
[30] V.Y.S. Lee, T.G. Karayiannis, Effect of inlet subcooling on flow boiling in mi- 234–252.
crochannels, Appl. Therm. Eng. 181 (2020). [51] M.P. David, J. Miler, J.E. Steinbrenner, Y. Yang, M. Touzelbaev, K.E. Goodson, Hy-
[31] I. Aranzabal, I.M. de Alegria, N. Delmonte, P. Cova, I. Kortabarria, Comparison draulic and thermal characteristics of a vapor venting two-phase microchannel
of the heat transfer capabilities of conventional single- and two-phase cooling heat exchanger, Int. J. Heat Mass Transf. 54 (25–26) (2011) 5504–5516.
systems for an electric vehicle IGBT power module, IEEE Trans. Power Electron. [52] H. Tuo, P. Hrnjak, New approach to improve performance by venting peri-
34 (5) (2019) 4185–4194. odic reverse vapor flow in microchannel evaporator, Int. J. Refrig. 36 (8) (2013)
[32] A. Bhunia, S. Chandrasekaran, C.L. Chen, Performance improvement of a power 2187–2195.
conversion module by liquid micro-jet impingement cooling, IEEE Trans. Com- [53] A. Mohiuddin, R. Loganathan, S. Gedupudi, Experimental investigation of flow
pon. Packag. Technol. 30 (2) (2007) 309–316. boiling in rectangular mini/micro-channels of different aspect ratios without
[33] A.S. Kousalya, C.N. Hunter, S.A. Putnam, T. Miller, T.S. Fisher, Photonically en- and with vapour venting membrane, Appl. Therm. Eng. 168 (2020) 114837.
hanced flow boiling in a channel coated with carbon nanotubes, Appl. Phys. [54] Y. Wang, Y. Peles, Subcooled flow boiling in a microchannel with a pin fin and
Lett. 100 (7) (2012). a liquid jet in crossflow, Int. J. Heat Mass Transf. 86 (2015) 165–173.
[34] D. Li, et al., Enhancing flow boiling heat transfer in microchannels for thermal [55] A. Radwan, S. Ookawara, M. Ahmed, Thermal management of concentrator
management with monolithically-integrated silicon nanowires, Nano Lett. 12 photovoltaic systems using two-phase flow boiling in double-layer microchan-
(7) (2012) 3385–3390 Jul 11. nel heat sinks, Appl. Energy 241 (2019) 404–419.
[35] T. Alam, W. Li, W. Chang, F. Yang, J. Khan, C. Li, A comparative study of flow [56] W. Li, J. Ma, T. Alam, F. Yang, J. Khan, C. Li, Flow boiling of HFE-7100 in silicon
boiling HFE-7100 in silicon nanowire and plainwall microchannels, Int. J. Heat microchannels integrated with multiple micro-nozzles and reentry micro-cav-
Mass Transf. 124 (2018) 829–840. ities, Int. J. Heat Mass Transf. 123 (2018) 354–366.
[36] F. Yang, W. Li, X. Dai, C. Li, Flow boiling heat transfer of HFE-70 0 0 in [57] W. Li, et al., Enhanced flow boiling in microchannels using auxiliary channels
nanowire-coated microchannels, Appl. Therm. Eng. 93 (2016) 260–268. and multiple micronozzles (I): characterizations of flow boiling heat transfer,
[37] Y.J. Choi, D.H. Kam, Y.H. Jeong, Analysis of CHF enhancement by magnetite Int. J. Heat Mass Transf. 116 (2018) 208–217.
nanoparticle deposition in the subcooled flow boiling region, Int. J. Heat Mass [58] W. Li, et al., Enhanced flow boiling in microchannels through integrating mul-
Transf. 109 (2017) 1191–1199. tiple micro-nozzles and reentry microcavities, Appl. Phys. Lett. 110 (1) (2017).
[38] A.K.M.M. Morshed, T.C. Paul, J.A. Khan, Effect of Al2O3 nanoparticle deposition [59] X. Jiang, S. Zhang, Y. Li, C. Pan, High performance heat sink with counter flow
on flow boiling performance of water in a microchannel, Exp. Therm. Fluid Sci. diverging microchannels, Int. J. Heat Mass Transf. 162 (2020).
47 (2013) 6–13. [60] D. Bogojevic, K. Sefiane, G. Duursma, A.J. Walton, Bubble dynamics and flow
[39] A.K.M.M. Morshed, T.C. Paul, J. Khan, Effect of Cu–Al2O3 nanocomposite coat- boiling instabilities in microchannels, Int. J. Heat Mass Transf. 58 (1–2) (2013)
ing on flow boiling performance of a microchannel, Appl. Therm. Eng. 51 (1–2) 663–675.
(2013) 1135–1143. [61] N. Patra, P. Ghosh, R.S. Singh, A. Nayak, Flow visualization in dilute oxide based
[40] A. Mukherjee, S.G. Kandlikar, The effect of inlet constriction on bubble growth nanofluid boiling, Int. J. Heat Mass Transf. 135 (2019) 331–344.
during flow boiling in microchannels, Int. J. Heat Mass Transf. 52 (21–22) [62] J.W. Palko, et al., Extreme two-phase cooling from laser-etched diamond and
(2009) 5204–5212. conformal, template-fabricated microporous copper, Adv. Funct. Mater. 27 (45)
[41] G. Wang, P. Cheng, A.E. Bergles, Effects of inlet/outlet configurations on flow (2017).
boiling instability in parallel microchannels, Int. J. Heat Mass Transf. 51 (9–10) [63] M.E. Steinke, S.G. Kandlikar, Single-phase liquid heat transfer in plain and
(2008) 2267–2281. enhanced microchannels, in: Proceedings of the International Conference on
[42] T. Zhang, J.T. Wen, Y. Peles, J. Catano, R. Zhou, M.K. Jensen, Two-phase refriger- Nanochannels, Microchannels, and Minichannels, 47608, 2006, pp. 943–951.
ant flow instability analysis and active control in transient electronics cooling [64] S. Kline, F. McClintock, Describing uncertainty in single sample experiments,
systems, Int. J. Multiph. Flow 37 (1) (2011) 84–97. Mech. Eng. 75 (1953) 3–8.
[43] O.K. Siddiqui, S.M. Zubair, Efficient energy utilization through proper design [65] Y. Tang, et al., Effects of structural parameter on flow boiling performance of
of microchannel heat exchanger manifolds: a comprehensive review, Renew. interconnected microchannel net, Appl. Therm. Eng. 112 (2017) 164–173.
Sustain. Energy Rev. 74 (2017) 969–1002. [66] S. Wang, H.H. Chen, C.L. Chen, Enhanced flow boiling in silicon nanowire–
[44] P.C. Lee, C. Pan, Boiling heat transfer and two-phase flow of water in a single coated manifold microchannels, Appl. Therm. Eng. 148 (2019) 1043–1057.
shallow microchannel with a uniform or diverging cross section, J. Micromech. [67] F. Yang, X. Dai, Y. Peles, P. Cheng, J. Khan, C. Li, Flow boiling phenomena in a
Microeng. 18 (2) (2008). single annular flow regime in microchannels (I): characterization of flow boil-
[45] B.R. Fu, P.H. Lin, M.S. Tsou, C. Pan, Flow pattern maps and transition criteria for ing heat transfer, Int. J. Heat Mass Transf. 68 (2014) 703–715.
flow boiling of binary mixtures in a diverging microchannel, Int. J. Heat Mass
Transf. 55 (5–6) (2011) 1754–1763.

15