International Journal of Heat and Mass Transfer 252 (2025) 127489

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: [Link]/locate/ijhmt

A microstructured counter-flow diverging microchannel heat sink for

cooling devices with ultra-high heat flux
Xingchi Jiang a,b, Syed Waqar Ali Shah a , Gong Chen a , Yuan Gao a , Yuanjie Li a , Enshen Long b,
Chin Pan a,*
a
Department of Mechanical Engineering, City University of Hong Kong, Kowloon, Hong Kong SAR, PR China
b
College of Architecture and Environment, Sichuan University, Chengdu 610065, PR China

A R T I C L E I N F O A B S T R A C T

Keywords: To address the critical challenge of thermal management for devices with ultra-high heat fluxes, this study ex­
Thermal management plores experimentally convective boiling heat transfer and two-phase flow for a counter-flow diverging micro­
Ultra-high heat flux channel (CFDM) heat sink with well-designed gradient microcavities and interlinked microgrooves. A counter-
Hybrid microstructures
flow configuration facilitates the channel-to-channel heat transfer to enhance bubble nucleation from the
Counter-flow microchannels
gradient cavities in the entrance region with subcooled liquid, while suppress void growth near the channel exit.
Two-phase flow
On the other hand, the capillary-driven flow through the centripetal microgrooves enables rapid liquid replen­
ishment for evaporation. Flow visualization confirms intensive nucleate boiling in the entrance region with
subcooled liquid. Experimental results demonstrate for a mass flux of 800 kg/m2s at the channel inlet an un­
precedented heat dissipation rate of 1041 W/cm² at 3 cm² area and the corresponding two-phase heat transfer
coefficient of 354 kW/m²K. Significantly, the two-phase flow is very stable with remarkably low pressure drop
within 10 kPa. The research findings reveal an effective cooling technology of counter-flow diverging micro­
channels combined with well-designed microcavities and microgrooves on the bottom surface. Such an inno­
vative design demonstrates great potential for the sustainable thermal management of high heat flux devices,
such as high-power electronics and the first wall of nuclear fusion reactors.

predicted that the cooling demand for the next generation of IGBT may
1. Introduction reach as high as 500 W/cm2 [4–6] or even exceed 1000 W/cm2 [7,8] in
the near future. Therefore, thermal management for power electronics
Until 2023, the carbon-neutral energy strategy has been proposed or demands technological breakthroughs with ultra-high heat flux dissi­
declared by over 120 countries worldwide [1,2]. Such a strategy will pation. Similarly, the nuclear fusion reactor, regarded as the ultimate
further promote the application of clean energy technology, such as the clean energy for mankind, is also facing ultra-high heat dissipation de­
electric vehicle (EV) revolution, battery energy storage system, photo­ mand. The first wall of a fusion reactor needs to be strong enough to
voltaic solar system, super data center for digitalization and artificial withstand the plasma and also be able to dissipate ultra-high heat flux
intelligence. Meanwhile, energy conversion and management among over 1000 W/cm2 [9,10]. Moreover, a large amount of water and elec­
these systems requires thermal management with ultra-high perfor­ tricity is consumed in supercomputing centers and hyperscale data
mance. Recent advances in thermal management, such as Ye and Arıcı’s centers for modern network cloud and artificial intelligence computing
numerical design for energy storage systems [3] demonstrates effective services [11,12]. Many real-world energy management applications
strategies for temperature uniformity and airflow optimization, urgently require high-performance cooling systems that can offer effi­
addressing critical challenges in modern energy systems under ciency, robustness, and reliability.
increasing power densities across EVs and photovoltaics, ensuring Traditional single-phase cooling has been developed for many de­
operational safety and stability. The power density of conversion elec­ cades by increasing the contact area of the solid-liquid interface and
tronics in EVs and photovoltaic solar panels keeps rising sharply. For enhancing the disturbance through the surface structures (i.e., rods, fins,
example, the heat dissipation demand of the current generation of IGBT spirals). However, limited by the heat transfer capacity per unit coolant
power chips in EVs has reached a power density of 250 W/cm2. It is mass flow rate, it is still challenging to meet the requirement for ultra-

* Corresponding author.
E-mail address: chinpan@[Link] (C. Pan).

[Link]
Received 14 March 2025; Received in revised form 19 May 2025; Accepted 30 June 2025
Available online 10 July 2025
0017-9310/© 2025 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
X. Jiang et al. International Journal of Heat and Mass Transfer 252 (2025) 127489

Nomenclature Tout Outlet temperature, ◦ C

Ts Heat sink temperature, ◦ C
Ab Base area of heat sink, m2 Tsat Saturation temperature, ◦ C
Ac Inlet cross-section area of all microchannels, m2 Tw Wall surface temperature, ◦ C
Cp Specific heat, J/kgK T∞ Bulk temperature, ◦ C
d Copper substrate thickness, m V̇ volume flow rate, m3/s
G Mass flux at the channel inlet, kg/m2s W Microchannel width, m
hfg Latent heat of vaporization, kJ/kg
hl Single-phase heat transfer coefficient, kW/m2K Greek symbols
htp Two-phase heat transfer coefficient, kW/m2K Δ Difference of parameters
kcu Thermal conductivity of copper, W/mK θ Contact angle, (degree)
ṁ Mass flow rate, kg/s δt Boundary layer thickness, m
P Average pressure, kPa ρv Vapor density, kg/m3
Pin Pressure at inlet, kPa σ Surface tension, N/m
Pout Pressure at outlet, kPa Abbreviations and acronyms
Qeff Heat transfer rate to fluid, W CHF Critical Heat Flux
Qloss Heat loss rate, W FDB Fully Developed Boiling
Qtotal Total heat input rate to heat sink, W HTC Heat Transfer Coefficient
qʹʹb Heat flux imposed on base area, W/cm2 ONB Onset of Nucleate Boiling
Ta Ambient temperature, ◦ C
Tin Inlet temperature, ◦ C

high power heat dissipation rate. Facing continuously integrated and alternating nucleate boiling with 1000 Hz frequency through these
miniaturized power electronics with high heat dissipation rate from a micro-slots can enhance the periodic liquid replenishment and sustain
small area, the flow boiling in microchannel has been widely regarded as thin film evaporation on the channel walls. Recinella and Kandlikar [21]
one of the most promising solutions [13–15]. The microchannel conducted a comparative experimental study between radial open
configuration offers a significantly larger heat transfer area, efficient microchannels and offset strip fins on the same base heating area of 1
utilization of latent heat from phase change, and increased flow turbu­ cm2, and a maximum heat flux of 897 W/cm2 was achieved by using
lence from bubble motion, which results in substantially higher heat radial offset strip fins. With more flow area for vapor volume expanding,
transfer capacity. Moreover, the wall temperature may be around the the offset strip fins also demonstrated better flow stability and lower
fluid boiling point and the temperature uniformity of the heating surface pressure drops. Later, Drummond et al. [22] designed a hierarchical
may be improved to some extent. manifold microchannel heat sink array on a 5 mm × 5 mm (0.25 cm2)
Despite numerous studies in the past few decades, the flow boiling silicon chip area, and achieved a high heat flux of up to 910 W/cm2 by
heat transfer in microchannels has not yet been commercialized on using HFE-7100 as the coolant. However, the two-phase pressure drop of
large-scale for cooling the power electronics. This may be due to some 162 kPa was quite high due to a very large mass flux of 2900 kg/m2s.
inherent shortcomings and complex heat transfer mechanisms of the Similarly, Woodcock et al. [23] proposed and investigated a Piranha
two-phase flow and heat transfer in microchannels. For example, the Pin-Fin (PPF) microstructures in the microchannel to enhance the flow
critical heat flux (CHF) is relatively low at low mass fluxes due to the boiling using HFE-7100. They achieved over 1000 W/cm2 heat flux on a
partial dryout of liquid film near the channel exit. The rapid void tiny heating area of 0.0395 cm2 with a very high mass flux of 3350
expansion also produces a sharp rise in two-phase flow pressure drop kg/m2s.
with heat flux and requiring large pumping power. The flow boiling in These above studies in the literature demonstrate ultra-high cooling
microchannels is intrinsically unstable due to bubble expansion in both performance up to 1000 W/cm2 for a relatively small heating area
forward and backward directions unless an inlet restriction, which in ranging from 0.04 cm2 to 1 cm2 and, for many cases, with relatively high
turn increases the flow resistance, is designed. Recent advancements mass fluxes. The heating areas employed in these studies are relatively
address these challenges. For example, Wang et al. [16] proposed small compared with the actual application area of power electronics, e.
tree-like microchannels via improved Murray’s law, achieving uniform g., IGBT chips of several to >10 cm2 [4,24,25]. Moreover, when
heat source coverage and a 30 % thermal resistance reduction, enabling approaching an ultra-high heat flux of about 1000 W/cm2, most studies
heat flux dissipation of 264 W/cm². Concurrently, Wang et al. [17] presented a huge two-phase flow pressure drop over the microchannel,
demonstrated pin-fin integration elevating the Nusselt numbers by 94.8 exceeding 100 kPa, due to very high mass flux and void fraction
%, albeit with increased flow resistance. These studies advance thermal development in the channel. The thin liquid film evaporates extremely
management by balancing heat transfer enhancement and structural fast under the ultra-high heat flux conditions and the explosively
optimization. increased void fraction will inevitably cause serious obstruction, espe­
Numerous scholars have devoted their efforts to improve the flow cially in the microchannels with very small hydraulic diameters.
boiling performance of the microchannel heat sink and achieve a high Without a well-established mechanism to control the void expansion, it
heat flux in a relatively small heat transfer area. For example, Mandel is difficult to achieve the ultra-high heat flux with stable two-phase flow
et al. [18] proposed a hermetically bonded embedded microchannel in microchannels while maintaining a low pressure drop.
manifold for two-phase cooling and achieved a high heat flux of 560 In authors’ previous study [26], a counter-flow diverging micro­
W/cm2. Later, they [19] further updated the integrated microchannel channel (CFDM) heat sink was developed configuration with an inno­
cooler and achieved an ultra-high heat flux of 1000 W/cm2 on a rela­ vative three-dimensional inlet manifold for a practical heat transfer area
tively small heating area of 1 cm2. Through microstructure design, of 3 cm × 4 cm. Such a novel design overcomes many inherent limita­
Huang et al. [20] achieved a CHF of 810 W/cm2 on a tiny heating area of tions of convective boiling in co-current microchannels, and achieves
0.2 cm2 by using micro-slots as nucleation sites; meanwhile, the excellent high heat transfer performance, very stable boiling two-phase

2
X. Jiang et al. International Journal of Heat and Mass Transfer 252 (2025) 127489

Fig. 1. (a) Assembly of test section with CFDM heat sink; (b) measurements of temperature and pressure; (c) top view of microchannel array.

flow and remarkably low two-phase pressure drop. Such high perfor­ 2. Experimental apparatus and methodology
mance is mainly due to the counter-flow and diverging microchannel
design, which enables significant channel-to-channel heat transfer ef­ 2.1. Experimental setup
fects suppressing the void growth near the channel exit, while activating
nucleate boiling toward the channel inlet. The performance can be Inheriting our previous work[26–28], the present study constructs a
significantly enhanced with micro- and nano-structures on the micro­ forced circulation loop system, which mainly includes a gear pump
channel bottom surface [27]. The addition of artificial nucleation sites (Fluid-o-Tech, FG300), flow meter (FLOMEC, OM004), filter, preheater,
and nano-structures on channel bottom surface significantly enhances test section, differential pressure transducer (Asmik, MIK-2051),
the performance with dissipating heat flux greater than 400 W/cm2 with condenser, liquid reservoir, and an auxiliary chiller. Deionized water
wall temperature limited to 140 ◦ C and without reaching the critical is selected as the working fluid in this study. Moreover, a DAQ system
heat flux. Moreover, the corresponding coefficient of performance, (National Instruments, PXIle-1082) with a customized NI LabVIEW
which is defined as the ratio of total heat dissipation rate to the pumping program is used for monitoring experimental parameters and data
power, is about 150,000, one order of magnitude higher than that re­ collection. Additionally, a high-speed camera (IDT, Crash Cam-1540) is
ported in the literature. set above the test section, and a frame rate of 5000 frames/s is adopted
The main novelty of the present study compared to our previous for flow visualization and recording the flow images for the analyses of
work [27] is the well-designed microgrooves and gradient microcavities. two-phase heat transfer mechanisms.
The former provides rapid liquid replenishment for thin-film evapora­ Fig. 1(a) illustrates the core test section, which is composed of the
tion, while the later considers the local subcooling to design the cavity CFDM heat sink (pure copper), a housing base (aluminum alloy), ther­
size to activate bubbles at different axial locations. Notably, the mal isolation plate (PEEK), distribution glass (quartz glass), inlet
laser-etched microgrooves and microcavities in the present study exhibit chamber layer plate (PEEK), visualization window (tempered glass), and
superior mechanical stability, while the nano-coated surfaces employed top cover frame (aluminum alloy). The housing base is fixed upon the
in our previous study [27] may be prone to delamination. heating module (as shown in Supporting Information S1) by screws to
The present study investigates experimentally the flow boiling per­ support the whole test section. The thermal isolation plate tightens the
formance of a CFDM heat sink with well-designed microcavities and a CFDM heat sink and provides reliable sealing. The distribution glass
practical heating area of 2 cm × 1.5 cm (3 cm2). The dynamic two-phase integrated with the plate for inlet chamber layer provides the counter-
flow images obtained using high-speed video camera are analyzed to flow microchannel array with proper inlet/outlet arrangement and re­
identify the Onset of Nucleate Boiling (ONB) location in the micro­ alizes the independent supply of subcooled liquid to every single
channels for a specific heat flux. The ONB location versus heat flux is microchannel of the CFDM heat sink. The glass provides a window for
compared with the subcooled entrance length based on energy balance two-phase flow visualization using a high-speed camera. The top cover
and a well-established correlation for ONB location to demonstrate the frame has set to screw all components tight with bolts and nuts. A high-
significant merits of the present design. The flow visualization also helps temperature resistant glue is adopted to seal the crevice between layers
to explain the very low two-phase flow pressure drop offered by the to prevent possible deformation of components under the conditions
current design with very stable two-phase flow at ultra-high heat fluxes. with ultra-high heat fluxes.
The present study harmonizes microscale mechanisms with system-level Fig. 1(b) shows the temperature measurement positions of the inlet/
performance offering significant advances toward cooling for devices outlet coolant and heat sink central location (3.5 mm below the
with ultra-high heat flux dissipation. microchannel bottom surface), and the probes of the differential pres­
sure transducer connecting to the inlet chamber and outlet chamber,
respectively. The temperature at the central location reflects the highest

3
X. Jiang et al. International Journal of Heat and Mass Transfer 252 (2025) 127489

wall temperature [23] and provides a wall superheat for conservative Table 1
evaluation of heat transfer coefficient. Uncertainties of major parameters.
Detailed dimensions of the microchannel array on the heat sink are Main Parameters Uncertainties Main Parameters Uncertainties
illustrated in Fig. 1(c). The 16 microchannels in total with counter-flow
Volume flow rate V̇ Mass flux G
design are arranged on a 20 mm × 15 mm base area. Each microchannel ±1.0 % ±4.18 %
wall has a diverging angle of 0.5◦ to form a diverging microchannel with Heating power ±1.0 % Base heat flux qb” ±5.84 %
the inlet width of 0.5 mm and outlet width of 0.85 mm. The depth of Qtotal
Temperature T ±0.40 % Wall superheat Tw-Tsat ±5.88 %
each microchannel maintains a constant value of 2 mm. The thickness of Pressure drop ΔP ±0.10 % Heat transfer coefficient ±8.28 %
the wall between two neighboring channels is 0.2 mm. htp
Channel dimension ±0.02 mm Coefficient of ±5.92 %
performance
2.2. Test procedure

The experiments are conducted based on the following procedure. 2.4. Uncertainty analysis
The test module is first assembled onto the heating block and integrated
into the flow loop system. Deionized water, serving as the working fluid, Uncertainties in the present study mainly come from machining
is preconditioned by degassing through vigorous boiling in the reservoir uncertainties of microchannel samples and measurements of heating
for over one hour to eliminate non-condensable gases. The inlet tem­ power, flow rate, temperature, and pressure drop. The uncertainty of the
perature is precisely regulated using a water-bath preheater, while the mechanical machining for the CFDM heat sink is ± 0.02 mm, provided
volumetric flow rate is controlled via a gear pump. Prior to testing, the by the manufacturer. The volume flow rate, heating power, and pressure
system is stabilized to ensure a steady state or quasi-steady state is drop are directly measured by a volumetric flow meter, a digital power
reached over 3 min. During experiments, temperatures and pressures are meter, and a differential pressure transducer, respectively. The tem­
recorded using a data acquisition (DAQ) system, with a sampling fre­ perature values related to calculations are measured by T-type ther­
quency of 1 Hz for 60 s per dataset. Synchronously, high-speed imaging mocouples, which yield measurement uncertainty of ± 0.5 ◦ C (in the
at 5000 frames per second captures two-phase flow patterns within the range of − 40~125 ◦ C) and ± 0.4 % (in the range of 125~350 ◦ C). Based
microchannels, enabling correlation between thermal-hydraulic per­ on the uncertainty propagation and analysis method proposed by Kline
formance and flow regime dynamics. Post-test validation includes [29], the uncertainties of main parameters can be evaluated as listed in
recalibration of sensors and verification of data consistency to minimize Table 1.
measurement uncertainties. This systematic approach ensures repro­
ducibility and aligns with established standards for two-phase flow 3. Microstructure design
experimentation.
3.1. Microgrooves and liquid spreading enhancement
2.3. Data reduction
Based on many research works in the literature [13,30–32] and our
The effective heating power is the total input power measured by a observations in the previous studies [27,33], annular flow usually ap­
power meter minus the estimated heat loss rate, which evaluation has pears in the middle to downstream region of the microchannels while
been illustrated in detail in Supporting Information S2. Therefore, the fully developed boiling (FDB) regime prevails and thin film boiling
effective base heat flux is calculated as: dominates the two-phase heat transfer. Under low heat flux conditions,
the liquid film may maintain a stable evaporation rate with continuous
Qeff replenishment of the upstream liquid for a given flow rate. However,
qʹʹb = (1)
Ab when an ultra-high heat flux is imposed, the rapid evaporation and
Qeff denotes the effective heating power absorbed by the working explosive vapor volume expansion in both forward and backward di­
fluid and Ab is the microchannel substrate area, which is 20 mm × 15 rections may hinder the liquid replenishment from upstream. The local
mm for the present study. When the test section achieves a steady state, dry-out spots may appear at the microchannel bottom surface due to the
the local wall temperature of microchannel bottom Tw is extrapolated high surface temperature as shown in Fig. 2(a), while the corners and
based on the one-dimension heat conduction as: side walls can still maintain a wetting state by the liquid film for a short
moment before the liquid is totally evaporated. With a proper design,
qʹʹb d the mainstream liquid or the liquid film attached to side corners may
Tw = Ts − (2)
kc u provide liquid replenishment to prevent the formation of local dry-out
Ts is the heat sink temperature measured by the T-type thermocouple patch in the liquid film between the vapor flow and the bottom sur­
embedded at the central location, as shown in Fig. 1(b). Where d denotes face and, therefore, enhances the heat transfer performance.
the distance from the thermocouple location to the channel bottom Microgrooves have demonstrated good wicking performance on the
surface (3.5 mm), kcu is the thermal conductivity of copper (390 W/mK). surface in the literature [34–37]. The present study proposes a novel
With the base heat flux given by Eq. (1), the two-phase heat transfer design of microgroove array to enable the microchannel surface with
coefficient (HTC) can be calculated based on Newton’s law of cooling as: good wicking performance, as shown in Fig. 2(b). The microgroove with
40 μm in width and 400 μm in depth is selected and arranged on the
qʹʹb channel bottom surface, and there are four microgrooves in the axial
htp = (3)
Tw − Tsat direction. Notably, the side and the middle grooves are connected by 45◦
Notably, the present study adopts the central wall temperature ac­ centripetal short grooves with the same width. Such a design of micro­
quired by Eq. (2) for a conservative evaluation of two-phase HTC. groove array can not only accelerate the liquid spreading from channel
The mass flux G at the channel inlet is adopted in the following upstream to downstream, but also replenish the liquid from the side
discussion, which can be readily evaluated by definition: corners to the central grooves timely to effectively prevent the occur­
rence of premature dry-out. More details of microgroove dimension
G = ṁ/Ac = V̇ ρ/Ac (4) selection and wicking performance tests are demonstrated in Supple­
mentary Information S3.
where ṁ denotes the mass flow rate, Ac is the inlet cross-section area of
all channels, and ρ is density of the DI-water at inlet.

4
X. Jiang et al. International Journal of Heat and Mass Transfer 252 (2025) 127489

Fig. 2. Liquid replenishment and spreading mechanisms in microchannels (a) Local dryout patch mitigation via upstream flow; (b) Centripetal microgrooves
enabling multi-directional liquid spreading.

3.2. Microcavities and preheating length for nucleate boiling [ ( )0.463P0.0234 ]

GAc CP qʹʹb qʹʹb
LONB = (Tsat − Tin ) − + 0.556 (6)
The fluid entering into the microchannel is usually at a certain qʹʹb W hl 1082P1.156
subcooling. The liquid temperature is increasing while flowing and
being heated in the microchannel until the saturated temperature is Where hl is the single-phase HTC before ONB, and P is the system
reached. The distance from the entrance to the location where the average pressure. The very existence of LONB for single-phase convection
saturation temperature is reached is referred to as the subcooled will inevitably result in a low average HTC over the microchannel. Eq.
entrance length. Based on the energy conservation, the subcooled length (6) suggests that for a given mass flux the higher the heat flux the shorter
can be readily derived as: the preheating length for nucleate boiling. As pointed out earlier, the
counter-flow design in a CFDM heat sink may provide extra heating to
GAc CP (Tsat − Tin ) the subcooled liquid in the entrance region through the channel-to-
Lsc = (5)
qʹʹb W channel heat transfer effect. Therefore, in a CFDM heat sink, the pre­
Bubble nucleation may take place when the wall temperature is heating length for nucleate boiling will be shorter.
higher than the saturation temperature for certain amount. Therefore, Such a shortening effect on preheating length for nucleate boiling
there is some heating distance from the channel inlet for the wall tem­ may be further enhanced with the introduction of optimal-sized cavities
perature to develop to meet the requirement for bubble nucleation. on the channel bottom wall. Cavities on the heating surface have been
Based on the Bergles and Rohsenow correlation (1963) for onset of proven very effective in initiating bubble nucleation at a lower wall
nucleate boiling (ONB) of water in a channel, the location for the ONB superheat in many previous researches [38–42]. Many scholars pro­
(LONB), i.e., the preheating length for nucleate boiling, can be predicted posed different schemes to form surface cavities with a uniform size [43,
by the following equation: 44] or random sizes [45,46] on microchannel surfaces. However, the
optimal cavity size may change, as the bulk fluid temperature in the

Fig. 3. Hybrid array of laser-etched microgrooves and gradient-diameter microcavities in a microchannel and local SEM images.

5
X. Jiang et al. International Journal of Heat and Mass Transfer 252 (2025) 127489

Fig. 4. CFDM heat sink with (a) bare surface (BS) and (b) a hybrid array of microgrooves and microcavities (GC), and corresponding SEM images.

microchannel gradually increases from the upstream to the downstream

section. This may significantly affect the minimum wall temperature 1200
demanded for nucleate boiling initiation. According to Hsu’s classical BS-G=400 kg/m2s Boiling Curves
model [47], the active cavity size ranges between the minimum and
BS-G=600 kg/m2s
BS-G=800 kg/m2s a-4
maximum diameters given by Eq. (7). 1000 GC-G=400 kg/m2s
GC-G=600 kg/m2s
( )[
δt sinθ Tw − Tsat GC-G=800 kg/m2s
dmin , dmax = 1
2(1 + cosθ) Tw − T∞ a-3
b-3
Base Heat Flux qb” (W/cm2)

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ] 800
8(1 + cosθ)(Tw − T∞ )Tsat σ
∓ 1− (7)
ρv hfg δt (Tw − Tsat )2
a-2
Therefore, the optimal cavity size will vary with bulk temperature. 600
b-2
This suggests that a unique array of cavities with gradient diameter
evaluated based on Eq. (7) could more effectively initiate bubble
nucleation along the entire length of flow directions. 400 a-1
For the present study with DI-water under atmospheric conditions as
the working fluid and the heat sink being made of copper, the corre­ b-1
sponding contact angle (θ) is 86◦ , the inlet temperature is set as a con­
200
stant of 70 ◦ C and the half of the microchannel inlet width of 0.5 mm is
selected as the boundary layer thickness. For the present study, the bulk back flow
ONB
temperature increases from 70 ◦ C at the inlet to the saturation temper­
paral dryout

ature of water at atmospheric pressure, i.e., 100 ◦ C, the optimal diam­ 0

-10 0 10 20 30 40 50 60
eter will gradually enlarge, i.e., from 5 μm to over 100 μm. Therefore,
Tw-Tsat (oC)
this study adopts an artificial laser-drilled microcavity array with
gradient diameter along the flow direction, assuming fluid temperature Fig. 5. Effects of mass flux on boiling curves for GC-CFDM and BS-CFDM
varies linearly from inlet temperature at the channel inlet to saturation samples at Tin=70 ◦ C.
temperature at the exit, with cavity diameter of 5 μm at the inlet and 100
μm at the exit, as illustrated in Fig. 3 for the microchannel with the The finalized counter-flow diverging microchannel heat sink was
hybrid microstructure design. assembled for flow boiling experiments. The detailed fabrication
method for etching microgrooves and microcavities is provided in
Supplementary Information S4. Fig. 4(a) shows the CFDM heat sink with
3.3. Innovative hybrid microstructure
bare surface, referred to as BS sample, and SEM image for the bottom
surface. Fig. 4(b) illustrates the CFDM sample with microgrooves and
A highly efficient heat transfer mode of two-phase flow in a channel
optimal-sized cavities, referred to as GC sample, and SEM image for the
requires active nucleate boiling with a very short preheating length in
microstructure on the bottom surface.
the upstream section. Meanwhile, the liquid film evaporation in the
downstream section can maintain stable liquid replenishment without
4. Results and discussion
local dry-out. Based on this concept, this study designs and implements a
hybrid microstructure with microgrooves and microcavity array with
4.1. Ultra-high heat transfer performance
gradient diameter over the full channel length, as shown in Fig. 3, to
achieve the goals of very short preheating length for ONB in the up­
Fig. 5 illustrates the effects of mass flux at the channel inlet on
stream region and with stable liquid replenishment for the downstream
boiling curve for both GC-CFDM and BS-CFDM samples. For the single-
region.
phase convection, there is insignificant difference between BS sample
To fabricate the microgrooves and microcavities, a nanosecond laser
and GC sample. The large difference between BS sample and GC sample
etcher with visual positioning (Raycus Laser, RFL-P30) was employed to
at G = 800 kg/m²s in the single-phase regime arises from their inherent
etch pre-designed groove/cavity arrays by AutoCAD onto microchannel
structural differences in single-phase convective heat transfer. At lower
surfaces. The optimal etching depth was determined via wicking tests.

6
X. Jiang et al. International Journal of Heat and Mass Transfer 252 (2025) 127489

Fig. 6. Two-phase flow patterns at various heat fluxes for a pair of channels (channel No 8 and No 9) in GC and BS CFDM samples.

sizes, the GC sample demonstrates much lower wall superheat for ONB,
i.e., of 7.4 ◦ C, 13.1 ◦ C, and 16.8 ◦ C, respectively, for G = 400, 600, and
800 kg/m2s compared with that ΔTsat=14.7 ◦ C, 18.4 ◦ C, and 22.7 ◦ C,
respectively, for the BS sample, as marked in the figure. As one may
expect, bubble nucleation in the heat sink is initiated at a lower wall
superheat for a lower flow rate. This may result in heat transfer per­
formance for low flow rates being approximately the same as for high
mass flux at low heat fluxes. For example, for the GC sample the boiling
curves for the three flow rates mingle with each other for heat flux
around 300 W/cm2.
The GC-CFDM sample demonstrates much better HTC for boiling
heat transfer than that for the BS-CFDM one. For the same heat flux, the
GC-CFDM sample presents significantly steeper boiling curves for the
three flow rates, locating much further left than the BS sample. This is
particularly the case with the highest mass flux of 800 kg/m2s of the
present study. With the well-designed cavities and the channel-to-
channel heat transfer effects, more nucleation sites are activated to
generate bubbles and nucleate boiling prevails along the full micro­
channel, as will be demonstrated in two-phase flow images in Fig. 6 of
Fig. 7. Comparison of subcooled entrance length, predicted ONB location by the next subsection for the data marked in Fig. 7.
Bergles & Rohsenow correlation, the measured ONB locations in BS and
The GC-CFDM sample exhibits superior performance in dissipating
GC samples.
an ultra-high heat flux of 1041 W/cm2 with a wall superheat of 29.4 ◦ C
for G = 800 kg/m2s without reaching the CHF. Such a high heat dissi­
mass fluxes (e.g., G = 400 kg/m²s, Re = 668), laminar flow prevails and pation rate can be attributed to the well-design micro-grooves spreading
the effects of surface microstructure on convective heat transfer is the liquid toward various nucleation cavities and the channel-to-channel
insignificant. However, at G = 800 kg/m²s, the Reynolds number is heat transfer effects via counter-flow design. Indeed, the chilling effect
1336. This may trigger turbulent flow in a microchannel. The critical that may condense part of the vapor near the channel outlet preventing
Reynolds number for transition from laminar flow to turbulent flow is pre-mature dryout of the liquid film. On the other hand, the heating
significantly lower for flow in a microchannel [48,49]. The gradient effect from the neighboring channels together with heat imposed from
microcavities and microgrooves in the GC sample amplify the turbulent the bottom surface activates nucleate boiling near the channel inlet with
effects resulting in about 51 % higher single-phase convective heat highly subcooled liquid. Consequently, an ultra-high heat flux is dissi­
transfer coefficient than that for the BS sample for the same mass flux. pated with corresponding HTC as high as 354 kW/m2K.
These deviations highlight the importance of geometric and surface Significantly, the BS-CFDM sample also demonstrates very high Heat
conditions in microscale flow regime prediction. Flux Before Deteriorating (HFBD), i.e., the flow become unstable and
The figure clearly indicates that the GC-CFDM sample demonstrates eventually partial dryout, of 806 W/cm2, 627 W/cm2 and 496 W/cm2
significantly better performance in all aspects of boiling heat transfer, i. for the mass fluxes of 800 kg/m2s, 600 kg/m2s and 400 kg/m2s,
e., ONB, HTC and CHF. With the well-design cavities having optimal

7
X. Jiang et al. International Journal of Heat and Mass Transfer 252 (2025) 127489

mass fluxes are much higher than the maximum heat fluxes corre­
10 sponding to wall temperature limit of about 150 ◦ C reported in our
Pressure Drop
9 pioneer study [28] for the BS sample with a larger heat transfer area of
12 cm2. Such a high HFBD demonstrates the significant advantages of
8 34.9% CFDM design enabling the channel-to-channel heat transfer effects,
which enhances HFBD significantly.
Pressure Drop ΔP (kPa)

7
Notably, such an ultra-high heat flux for GC-CFDM sample leading to
6 a base temperature exceeding the high-temperature tolerance of the
22.7% heating model. The highest heat flux of 1041 W/cm2 was measured
5 before the heating base was melted. Notably, in our experiments, the
4 coolant inlet temperature was rigorously maintained at 70 ◦ C. For the
18.7% present case with the highest heat flux of 1041 W/cm2, the fluid outlet
3 temperature reaches the saturation temperature (115.8 ◦ C) corre­
G=400,Tin=70
BS-G=400 kg/m2s
G=600,Tin=70
BS-G=600 kg/m2s
sponding to the system pressure of about 1.8 bar, i.e., the average of the
2
G=800,Tin=70
BS-G=800 kg/m2s inlet and out pressure. Significantly, the GC-CFDM sample still demon­
GC-G=400,Tin=70
GC-G=400 kg/m2s
strates a vigorous and stable two-phase flow under such a harsh condi­
1 GC-G=600,Tin=70
GC-G=600 kg/m2s
GC-G=800,Tin=70
GC-G=800 kg/m2s tion for the heating block indicating that the CHF can be even higher.
0 The video recorded by the high-speed camera shows that only 1 or 2
0 200 400 600 800 1000 1200 microchannels are accompanied by minor instability as demonstrated in
Base Heat Flux qb” (W/cm2)
Supplementary Video SV-3. Significantly, such a high heat flux dissi­
pated is for a surface with base area of 2 cm × 1.5 cm, i.e., 3 cm2, which
Fig. 8. Effects of mass flux on pressure drops for GC and BS samples at
Tin=70 ◦ C. is very useful for practical applications.

4.1.1. Two-phase flow pattern

respectively. Unlike CHF, HFBD does not lead to sharp temperature rise,
To reveal the mechanisms underlying the ultra-high heat transfer
but to a state with unstable two-phase flow. Although the CHF is not
performance for GC and BS CFDM samples, the corresponding boiling
immediately triggered, insufficient capillary-driven liquid replenish­
two-phase flow patterns at various heat fluxes are examined for mass
ment due to lack of the microgrooves and insufficient channel-to-
flux of 800 kg/m2s. The high-speed images in Fig. 6 illustrate the two-
channel heat transfer effects caused by insufficient nucleate boiling
phase flow patterns for a pair of CFDM channels, i.e., channel No 8
due to lack of microcavities induces intermittent vapor blockage and
and channel No 9, corresponding to the data marked in the boiling curve
liquid starvation. This results in oscillatory two-phase flow (please refer
of GC and BS samples with G = 800 kg/m2s (see Fig. 5). As for the GC-
to Fig. 8(b) later in Section 4.3) leading to significant wall temperature
CFDM sample, the image Fig. 6 (a-1) illustrates the tiny bubbles
fluctuations (as shown in Fig. 9(a) later in Section 4.3). It can be seen
dispersed in the channels for a low heat flux of about 330 W/cm2. Since
that the limit to which stable heat dissipation can be achieved is
the bubble frequency is extremely high, the high-speed video camera
significantly affected by the flow rate, and more explanation can be
may not catch a clear image of bubbles. The schematic diagram shows
found in the Supporting Information S5. These HFBDs for the two lower

Fig. 9. Evolvement of two-phase flow patterns for the microchannel No 8 of (a) GC sample and (b) BS sample.

8
X. Jiang et al. International Journal of Heat and Mass Transfer 252 (2025) 127489

the contour of various bubbles and annular flow in the channel with the 4.1.2. ONB location versus heat flux
help of video to identify the sites with bubble nucleation and departure. To evaluate the significant contributions to boiling incipience by the
Fig. 6 (a-1) also indicates that most microcavities function as designed counter-flow configuration and the well-designed microcavities, the
and some bubbles may merge. In contrast, for approximately the same subcooled entrance length given by Eq. (5) and the preheating length for
heat flux, only one active nucleation site can be identified in the the first bubble nucleation in the channel, i.e., the ONB location pre­
downstream part of channel 8 and channel 9 in the BS-CFDM sample, as dicted by Eq. (6) versus heat flux are compared with the experimental
shown schematically in Fig. 6 (b-1). data identified from flow visualization, as for some of the heat fluxes
When the heat flux is increased to a medium level of about 608 W/ shown in Fig. 6, under the condition of G = 800 kg/m2s and Tin=70 ◦ C.
cm2, nucleate boiling prevails with tiny bubbles in the GC sample ex­ The blue curve in Fig. 7 shows the subcooled entrance length versus heat
tends much further to the upstream section in the GC-CFDM sample, as flux. It indicates that for low heat fluxes a very long distance is required
shown in the image Fig. 6 (a-2). The preheating length for triggering the to heat up the subcooled water at 70 ◦ C to become saturated. For heat
ONB in the channel is only about 20 % of the full microchannel length. flux below about 300 W/cm2 the heating length is actually longer than
Very active nucleate boiling prevails in the downstream 80 % channel the channel length of 20 mm and the water in the whole channel is
length and demonstrates a highly effective two-phase heat transfer subcooled. The green curve demonstrates the ONB location versus heat
mechanism. Indeed, for such a heat flux of about 600 W/cm2, the wall flux predicted by Bergles and Rohsenow (B&R) correlation. As one may
superheat for GC sample is 21.8 ◦ C only. However, a half-length of the expect, the preheating length for bubble nucleation is relatively smaller
microchannel in BS sample is occupied by single-phase liquid due to the than the subcooled entrance length indicating the subcooled flow
relatively low inlet temperature of Tin=70 ◦ C set in this study, more boiling in the channel. The experimental data shown by purple triangle
heating energy is needed to trigger the ONB and boiling heat transfer is symbols for the BS-CFDM sample demonstrate that the channel-to-
delayed to some extent, as shown in Fig. 6 (b-2). The corresponding wall channel heat transfer effect owing to the counter-flow design provides
superheat is 37.6 ◦ C, much higher than that in GC-sample. an extra heating effect in the entrance region to activate bubbles and
When a high heat flux of 813 W/cm2 is imposed, the GC sample significantly reduce the pre-heating length comparing with the B&R
shows a very unique flow pattern that the tiny bubbles appear right predictions. Errors are calculated in high-speed image measurements
under the inlet holes, as marked with red frames in image Fig. 6 (a-3). and shown as the error bars.
This demonstrates nucleate boiling and convective evaporation pre­ The curve with the red solid diamond symbols indicates the experi­
vailing over the whole channel, even though the inlet temperature is as mental data of the ONB locations for the GC-CFDM sample. The data
low as 70 ◦ C. Significantly, the two-phase flow pattern along the shown in the figure are the mean of the ONB locations among all 16
microchannel exhibits a very uniform and stable state. This results in a channels. The figure clearly demonstrates that the heating length for
relatively low wall superheat of 24.3 ◦ C. The corresponding high-speed ONB location for the GC-CFDM sample is much shorter than both the
video for the GC-CFDM sample at such a high heat flux can be seen in data for the BS-CFDM sample and the predictions of B&R correlation.
Supplementary Video SV-2. For approximately the same heat flux, This can be attributed to the unique counter-flow diverging micro­
nucleate boiling and annular flow heat transfer extend to about 60 % of channel configuration and the well-designed microcavities in the GC-
channel 8 and channel 9 in the BS-CFDM sample, demonstrating a CFDM sample, particularly for the low to medium heat flux region
relatively poor ONB performance on the bare copper surface, as shown from 200 to 600 W/cm2. Most of the data indicates that the preheating
in Fig. 6 (b-3) and Supplementary Video SV-1. As indicated earlier, for length for first bubble nucleation in the channel is smaller than 5 mm
such a high heat flux, the boiling two-phase flow becomes unstable and from entrance, i.e., 25 % of channel length.
partial dryout may appear intermittently. The heat transfer deteriorates Overall, attributing to the channel-to-channel heat transfer effect by
and the corresponding superheat is as high as 40.6 ◦ C. the counter-flow design provides extra heating effect to activate bubbles
The GC-CFDM sample can dissipate even higher heat flux. Indeed, from the microcavities designed, a considerable reduction in the pre­
Fig. 6 (a-4) demonstrates a very stable annular flow for the case with the heating length for the ONB location and an outstanding heat transfer
highest heat flux of 1041 W/cm2. Again, one can see bubble nucleation performance are brought by such a novel GC-CFDM design.
starts right under the inlet holes and nucleate boiling and liquid film
evaporation prevails throughout the whole channel. The corresponding 4.2. Very low two-phase pressure drop
video can be seen in Supplementary Video SV-3. From the video, one can
also identify vapor stem demonstrating nucleate boiling under the liquid Fig. 8 illustrates the pressure drop versus heat flux for the three flow
film. Such an ultra-high heat flux for the GC-CFDM sample is accom­ rates investigated in the present study. The BS-CFDM sample presents
plished by the combination of two innovative designs: CFDM configu­ relatively mild increment in two-phase pressure drop for the different
ration and surface microstructures. As indicated earlier, such a mass fluxes in this study. The diverging design with increasing flow area
microstructure combines the microgrooves providing the wicking effect to reduce flow velocity in the flow direction will counteract with the
for liquid spreading effectively to the downstream surface for thin film acceleration effects due to vaporization. Moreover, the high flow ve­
evaporation and nucleate boiling from the gradient-diameter cavity locity near the channel inlet will enhance the convective heat transfer of
array warrant excellent heat transfer throughout the whole micro­ the cool side near the channel inlet and, therefore, improve the channel-
channel. The CFDM configuration enables channel-to-channel heat to-channel heat transfer. On the other hand, the temperature difference
transfer, which activates nucleate boiling in the entrance region and between the subcooled inlet flow and saturated two-phase mixture near
suppressing void growth and stabilizing the two-phase flow. the channel outlet empowers a unique neighboring chilling effect
Interestingly, the outlet vapor quality for the case with the highest resulting in partial condensation and void reduction in the downstream
heat flux for the GC-CFDM sample at G = 800 kg/m²s and q”=1041 W/ section. This reduces the frictional pressure drop as well as the accel­
cm2 is 2.42 % based on the energy balance calculation using the satu­ eration pressure drop. Therefore, the two-phase flow pressure drop may
ration properties corresponding to the system pressure of 1.8 bar. This not significantly increase with increasing heat flux, as shown in the
will create a very large void fraction at low system pressure at the figure.
channel outlet. However, Fig. 6 (a-4) illustrates a very uniform void The GC-CFDM sample presents higher pressure drops, for both
fraction distribution along the whole channel. This must attribute to the single-phase and two-phase regimes, than that of the BS sample for all
channel-to-channel heat transfer effect, which will suppress void growth mass fluxes. For the highest heat flux of BS-CFDM sample, the two-phase
near the channel outlet, while promote nucleate boiling in the entrance pressure drops of the GC-CFDM sample are 18.7 % to 34.9 % higher. This
region of adjacent microchannels. is because the deep laser-etched grooves essentially lead to the micro­
channel surface being relatively rough, as shown previously in Fig. 4(b),

9
X. Jiang et al. International Journal of Heat and Mass Transfer 252 (2025) 127489

Fig. 10. Fluctuations of (a) wall temperature and (b) inlet pressure on GC and BS samples under G = 800 kg/m2s and a similar heat flux near 800 W/cm2.

and the flow resistance is, therefore, increased. Again, due to the channel-to-channel heat transfer, proposed earlier by the authors, with
counter-flow design, the elevation of pressure drop from the single- microgrooves and optimal-sized microcavities on the bottom wall of the
phase regime to the two-phase regime at the highest heat flux is also microchannels. The well-designed microgrooves and microcavities
mild with 53 %~84 % increase only. Remarkably, for the case with an together with counter-flow design enables intensive nucleate boiling in
ultra-high heat flux, i.e., qʹʹb =1041 W/cm2, the corresponding two-phase the entrance region with subcooled liquid and suppress void fraction
flow pressure drop is within 9.5 kPa. This shows another superior growth near the channel outlet. Consequently, nucleate boiling and
characteristic of GC-CFDM sample. convective evaporation prevail over the whole channel and an ultra-
high and unprecedented heat flux of 1041 W/cm2 with corresponding
HTC over 350 kW/m2K is demonstrated for a relatively low mass flow
4.3. Highly stable two-phase flow rate of 800 kg/m2s with the two-phase pressure drop below 10 kPa on a
practical heating area of 2.0 cm × 1.5 cm (3 cm2) and very stable boiling
The GC-CFDM sample demonstrates very stable two-phase flow even two-phase flow.
for ultra-high heat flux dissipation. To illustrate such a salient feature, The proposed counter-flow diverging microchannel with hybrid
Fig. 9(a) illustrates the evolution of two-phase flow pattern for micro­ microstructures on bottom wall is robust, easy to fabricate, and can be
channel No 8 for 14 ms randomly selected under the condition of qʹʹ = modularized or extended for applications with even larger areas. The
813 W/cm2 and G = 800 kg/m2s. The figure indicates that the location device may be employed effectively for the cooling of high-heat-flux
of the upstream edge of annular flow is very stable with very small components, such as power electronics and the first wall of a nuclear
fluctuations. Small bubbles are actively generated from the cavities at or fusion reactors with high energy efficiency.
very close to the inlet and expand rapidly and smoothly with heat
imposed from the heating base and heat transfer from the neighboring
CRediT authorship contribution statement
channels resulting in a steady annular flow to the downstream. There is
no sign of dry-out on the bottom surface downstream, as the chilling
Xingchi Jiang: Writing – original draft, Visualization, Validation,
effect from the neighboring channels condenses part of the vapor and the
Methodology, Investigation, Funding acquisition, Formal analysis, Data
excellent wicking effect through the interconnected microgrooves
curation, Conceptualization. Syed Waqar Ali Shah: Visualization.
effectively replenishes the liquid toward the channel exit. Therefore, the
Gong Chen: Investigation. Yuan Gao: Investigation. Yuanjie Li:
two-phase flow pattern can be maintained very stably, as shown in
Investigation. Enshen Long: Investigation. Chin Pan: Writing – review
Supplementary Video SV-2. Such stable two-phase flow may not be
& editing, Supervision, Methodology, Funding acquisition,
sustained in the BS-CFDM sample at the same heat flux. Significant
Conceptualization.
fluctuations in the upstream edge of bubble slug (as marked with green
line) and intermittent partial dryout (as marked with red rectangular
region) may be present for a heat flux of 806 W/cm2, as shown in Fig. 9 Declaration of competing interest
(b). This demonstrates another superior feature of the GC-CFDM sample.
The stable two-phase flow in the GC-CFDM sample also results in The authors declare that they have no known competing financial
very small fluctuations in the wall temperature and inlet pressure, as interests or personal relationships that could have appeared to influence
shown in Fig. 10. The figure indicates that both the instant wall tem­ the work reported in this paper.
perature and inlet pressure are very stable with very small fluctuations.
The fluctuations in wall temperature are within 1.0 ◦ C and within 2 kPa Acknowledgement
in inlet pressure. On the other hand, the fluctuations for BS-CFDM
sample are more dramatic. The fluctuations in wall temperature are The present study was sponsored by Internal projects 9380091 and
up to 6.2 ◦ C and up to 45 kPa in inlet pressure. This clearly demonstrates 7005233 of the City University of Hong Kong (CityU), National Natural
the significant merits of adding well-design microgrooves and micro­ Science Foundation of China (No. 52408121), Natural Science Foun­
cavities in a CFDM heat sink. dation of Sichuan Province (2024YFHZ0343) and partially supported by
a GRF project 9042991 of RGC of Hong Kong.
5. Conclusions
Supplementary materials
The present study proposes and implements an innovative hybrid
strategy for convective boiling in microchannels, combining the unique Supplementary material associated with this article can be found, in
manifold of counter-flow diverging microchannels enabling significant the online version, at doi:10.1016/[Link].2025.127489.

10
X. Jiang et al. International Journal of Heat and Mass Transfer 252 (2025) 127489

Data availability [24] A. Bhunia, S. Chandrasekaran, C.L. Chen, Performance improvement of a power
conversion module by Liquid Micro-jet impingement cooling, IEEE Trans. Compon.
Packag. Technol. 30 (2007) 309–316.
Data will be made available on request. [25] U.M. Choi, F. Blaabjerg, K.B. Lee, Study and handling methods of power IGBT
module failures in power electronic converter systems, IEEe Trans. Power.
References Electron. 30 (2015) 2517–2533.
[26] X. Jiang, S. Zhang, Y. Li, C. Pan, High performance heat sink with counter flow
diverging microchannels, Int. J. Heat. Mass Transf. 162 (2020) 120344.
[1] J.M. Chen, Carbon neutrality: toward a sustainable future, Innovation 2 (2021) [27] X. Jiang, [Link] Ali Shah, J. Liu, Y. Li, S. Zhang, Z. Wang, et al., Design of micro-
100127. nano structures for counter flow diverging microchannel heat sink with
[2] L. Chen, G. Msigwa, M. Yang, A.I. Osman, S. Fawzy, D.W. Rooney, et al., Strategies extraordinarily high energy efficiency, Appl. Therm. Eng. 209 (2022) 118229.
to achieve a carbon neutral society: a review, Env. Chem. Lett. 20 (2022) [28] X. Jiang, S. Zhang, Y. Li, Z. Wang, C. Pan, Achieving ultra-high coefficient of
2277–2310. performance of two-phase microchannel heat sink with uniform void fraction, Int.
[3] W.B.Y. Arıcı, Numerical thermal control design for applicability to a large-scale J. Heat. Mass Transf. 184 (2021) 122300.
high-capacity lithium-ion energy storage system subjected to forced cooling, [29] S.J. Kline, Describing uncertainty in single sample experiments, Mech. Eng. 75
Numer. Heat Transf. A 85 (2024). (1953) 3–8.
[4] P. Wang, P. McCluskey, A. Bar-Cohen, Two-phase liquid cooling for thermal [30] T. Chen, S.V. Garimella, Measurements and high-speed visualizations of flow
management of IGBT power electronic module, J. Electron. Packag. 135 (2013) boiling of a dielectric fluid in a silicon microchannel heat sink, Int. J. Multiph. Flow
021001. 32 (2006) 957–971.
[5] C. Qian, A.M. Gheitaghy, J. Fan, H. Tang, B. Sun, H. Ye, et al., Thermal [31] S.G. Kandlikar, Nucleation characteristics and stability considerations during flow
management on IGBT power electronic devices and modules, IEEe Access. 6 (2018) boiling in microchannels, Exp. Therm. Fluid. Sci. 30 (2006) 441–447.
12868–12884. [32] P. Thiangtham, C. Keepaiboon, P. Kiatpachai, L.G. Asirvatham, O. Mahian, A.
[6] I. Aranzabal, I.M. de Alegria, N. Delmonte, P. Cova, I. Kortabarria, Comparison of S. Dalkilic, et al., An experimental study on two-phase flow patterns and heat
the heat transfer capabilities of conventional single- and two-phase cooling systems transfer characteristics during boiling of R134a flowing through a multi-
for an electric vehicle IGBT power module, IEEe Trans. Power. Electron. 34 (2019) microchannel heat sink, Int. J. Heat. Mass Transf. 98 (2016) 390–400.
4185–4194. [33] K.H. Chang, C. Pan, Two-phase flow instability for boiling in a microchannel heat
[7] C. Zweben, Thermal materials solve power electronics challenges, Power Electron. sink, Int. J. Heat. Mass Transf. 50 (2007) 2078–2088.
Technol. 32 (2006) 40–47. [34] R. Fu, X. Hu, Y. Yan, W. Zhou, H. Wu, J. Wang, et al., Experimental study of the
[8] A. Datta, V. Sharma, D. Sanyal, P. Das, A conjugate heat transfer analysis of effects of nanofluids on wicking ability and thermal performance of a vertical open
performance for rectangular microchannel with trapezoidal cavities and ribs, Int. J. microgrooves heat sink, Int. J. Heat. Mass Transf. 144 (2019) 118674.
Therm. Sci. 138 (2019) 425–446. [35] L. Duan, Z. Wang, G. Chen, Y. Tang, Y. Sun, G. Zhong, et al., Capillary Wicking in
[9] F.L. Ribe, Fusion reactor systems, Rev. Mod. Phys. 47 (1975) 7–41. Double-Scale Composite Microgroove Wicks For Copper-Aluminum Composite
[10] Y. Wan, J. Li, Y. Liu, X. Wang, V. Chan, C. Chen, et al., Overview of the present Vapor Chambers, 126, International Communications in Heat and Mass Transfer,
progress and activities on the CFETR, Nucl. Fusion 57 (2017) 102009. 2021 105449.
[11] N. Jones, How to stop data centres from gobbling up the world’s electricity, Nature [36] A. Samanta, W. Huang, A. Parveg, P. Kotak, R.C.Y. Auyeung, N.A. Charipar, et al.,
561 (2018) 163–167. Enabling superhydrophobicity-guided superwicking in metal alloys via a
[12] R. van Erp, R. Soleimanzadeh, L. Nela, G. Kampitsis, E. Matioli, Co-designing nanosecond laser-based surface treatment method, ACS. Appl. Mater. Interfaces. 13
electronics with microfluidics for more sustainable cooling, Nature 585 (2020) (2021) 41209–41219.
211–216. [37] G. Jiang, Z. Tian, X. Luo, C. Chen, X. Hu, L. Wang, et al., Ultrathin aluminum wick
[13] I. Mudawar, Two-phase microchannel heat sinks: theory, applications, and with dual-scale microgrooves for enhanced capillary performance, Int. J. Heat.
limitations, J. Electron. Packag. 133 (2011) 041002. Mass Transf. 190 (2022) 122762.
[14] S.G. Kandlikar, History, advances, and challenges in liquid flow and flow boiling [38] A. Koşar, C.J. Kuo, Y. Peles, Boiling heat transfer in rectangular microchannels
heat transfer in microchannels: a critical review, J. Heat Transf. 134 (2012) with reentrant cavities, Int. J. Heat. Mass Transf. 48 (2005) 4867–4886.
034001. [39] C.J. Kuo, Y. Peles, Local measurement of flow boiling in structured surface
[15] N.H. Naqiuddin, L.H. Saw, M.C. Yew, F. Yusof, T.C. Ng, M.K. Yew, Overview of microchannels, Int. J. Heat. Mass Transf. 50 (2007) 4513–4526.
micro-channel design for high heat flux application, Renew. Sustain. Energy Rev. [40] S.G. Kandlikar, Scale effects on flow boiling heat transfer in microchannels: a
82 (2018) 901–914. fundamental perspective, Int. J. Therm. Sci. 49 (2010) 1073–1085.
[16] G. Wang, K. Sheng, Y. Wang, G. Ding, D. Xie, Thermal hydraulic performance of [41] J.Y. Lee, M.H. Kim, M. Kaviany, S.Y. Son, Bubble nucleation in microchannel flow
tree-like microchannel heat sink with high branching level based on the improved boiling using single artificial cavity, Int. J. Heat. Mass Transf. 54 (2011)
Murray’s law, Int. J. Heat. Mass Transf. 231 (2024) 125875. 5139–5148.
[17] G. Wang, Z. Wang, L. Lai, D. Xie, Y. Zhu, G. Ding, et al., Experimental and [42] S. Waqar Ali Shah, X. Jiang, Y.K.H. Li, G. Chen, J. Liu, Y. Gao, et al., Ultra-stable
numerical investigation of hydrothermal performance of a microchannel heat sink counter flow diverging minichannel heat sink with integrated microstructures for
with pin fins, Case Stud. Therm. Eng. 60 (2024) 104631. superior cooling performance, Appl. Therm. Eng. 258 (2025) 124560.
[18] R. Mandel, S. Dessiatoun, M. Ohadi, Embedded two-phase cooling of high flux [43] P.H. Lin, B.R. Fu, C. Pan, Critical heat flux on flow boiling of methanol–water
electronics using a directly bonded FEEDS manifold, in: 2016 15th IEEE mixtures in a diverging microchannel with artificial cavities, Int. J. Heat. Mass
Intersociety Conference on Thermal and Thermomechanical Phenomena in Transf. 54 (2011) 3156–3166.
Electronic Systems (ITherm), IEEE, 2016, pp. 77–84. [44] W. Li, X. Qu, T. Alam, F. Yang, W. Chang, J. Khan, et al., Enhanced flow boiling in
[19] R.K. Mandel, D.G. Bae, M.M. Ohadi, Embedded two-phase cooling of high flux microchannels through integrating multiple micro-nozzles and reentry
electronics via press-fit and bonded FEEDS coolers, J. Electron. Packag. 140 (2018) microcavities, Appl. Phys. Lett. 110 (2017) 014104.
031003. -1. [45] S. Li, R. Furberg, M.S. Toprak, B. Palm, M. Muhammed, Nature-inspired boiling
[20] G. Huang, W. Li, J. Ma, C. Ren, C. Li, High-frequency alternating nucleate boiling enhancement by novel nanostructured macroporous surfaces, Adv. Funct. Mater.
of water enabled by microslot arrays in microchannels, Int. J. Heat. Mass Transf. 18 (2008) 2215–2220.
150 (2020) 119271. [46] M.S. El-Genk, A.F. Ali, Enhanced nucleate boiling on copper micro-porous surfaces,
[21] A. Recinella, S.G. Kandlikar, Enhanced flow boiling using radial open Int. J. Multiph. Flow 36 (2010) 780–792.
microchannels with manifold and offset strip fins, J. Heat Transf. 140 (2018) [47] Y.Y. Hsu, On the size range of active nucleation cavities on a heating surface,
021502. -1. J. Heat Transf. 84 (1962) 207–213.
[22] K.P. Drummond, D. Back, M.D. Sinanis, D.B. Janes, D. Peroulis, J.A. Weibel, et al., [48] S.G. Kandlikar, in: Microchannels and Minichannels: Proceedings of the Third
A hierarchical manifold microchannel heat sink array for high-heat-flux two-phase International Conference, American Society of Mechanical Engineers, 2005.
cooling of electronics, Int. J. Heat. Mass Transf. 117 (2018) 319–330. [49] N.T. Obot, Toward a better understanding of friction and heat/mass transfer in
[23] C. Woodcock, C. Ng’oma, M. Sweet, Y. Wang, Y. Peles, J. Plawsky, Ultra-high heat microchannels — a literature review, Microscale Thermophys. Eng. 6 (2002)
flux dissipation with Piranha Pin Fins, Int. J. Heat. Mass Transf. 128 (2019) 155–173.
504–515.

11