University of South Carolina College of Engineering and Computing Mechanical Engineering Department

Enhanced flow boiling in microchannels

by self-sustained high frequency two-

phase oscillations

Fanghao Yang, Xianming Dai, Chih-Jung Kuo, Yoav Peles, Jamil Khan,

Chen Li

Enhanced flow boiling in microchannels by self-sustained high frequency

two-phase oscillations

Fanghao Yang a, Xianming Dai a, Chih-Jung Kuo b, Yoav Peles b, Jamil Khan a, Chen Li a,⇑

aDepartment of Mechanical Engineering, University of South Carolina, Columbia, SC 29210, United StatesbDepartment of Mechanical, Aerospace & Nuclear Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180, United States

a r t i c l e i n f o

Article history:

Received 16 June 2012

Received in revised form 3 November 2012

Accepted 14 November 2012

Keywords:

Enhanced flow boiling

Microchannels

High frequency

Self-sustained two-phase oscillation

a b s t r a c t

Experimental study of flow boiling heat transfer in a microchannel array consisting of main channels con-

nected to two auxiliary channels (each) was conducted. A microbubble-excited actuation mechanism,

powered by high frequency vapor bubble growth and collapse, was established to create and sustain

strong mixing in the microchannels. It was shown to significantly enhance flow boiling heat transfer

in microchannels. Experimental studies were conducted at mass fluxes ranged from 150 to 480 kg/

m2 s with de-ionized (DI) water as the working fluids. Compared with microchannels with inlet restric-

tors (IRs), the average two-phase heat transfer coefficient was improved by up to 149%. More importantly,

a 71–90% reduction in pressure drop at moderate mass fluxes ranged from 400 to 1400 kg/m2 s was

observed. Heat flux up to 552 W/cm2 at a mass flux of 480 kg/m2 s was demonstrated. Flow and heat

transfer mechanisms were studied and discussed.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Two-phase flow instabilities are challenging issues hindering

the practical implementations of flow boiling in micro domains.

Flow boiling instability in microchannels are linked to several

instability modes including, but not limited to, upstream volume

and parallel channel effects. If not properly suppressed, they can

lead to premature critical heat flux (CHF) condition at low exit va-

por quality [1–5]. Flow boiling instabilities are also classified into

two general categories: static and dynamic [3]. Most dynamic flow

instabilities are associated with two-phase oscillation (TPO) fre-

quencies that range from 0.06 to 2 Hz in the laminar flow regime

[1–5]. These TPOs are low-frequency and large-amplitude fluctua-

tions [2,6–9]. Thus, they can delay rewetting of dry patches formed

on the heat transfer surfaces, and consequently, trigger CHF at low

heat fluxes, i.e., premature CHF conditions. High frequency TPOs

(100 Hz) are most commonly achieved through active-controlled

microsecond heating pulses with significant sacrifice of volumetric

flow rates [2,10–14]. Recent studies demonstrated that novel con-

figurations, such as inlet restrictors (IRs) or valves/orifices, can sup-

press boiling instabilities and enhance several key flow boiling

parameters including onset of nucleate boiling (ONB), heat transfer

coefficient (HTC), and CHF [15–20]. Several studies have shown

that other configurations, such as impingement jets, can effectively

suppress flow reversal and at the same time enhance HTCs and

CHFs [7,21–23].

Although the aforementioned techniques successfully enhanced

nucleate boiling and suppressed flow boiling instabilities in micro-

scale systems, the additional pressure drop (Dp) and power

requirements introduced by these configurations are a major hin-

drance. For example, IRs or orifices can effectively suppress flow

instabilities and enhance CHF [24–27], but, they increase the pres-

sure drop by up to five-fold. High flow resistance can cause critical

issues, such as high pumping power and coolant leakage.

In this study, IRs were employed to trap bubble slugs inside

microchannels. Two auxiliary channels were connected to the

main channel through a cross junction to induce auxiliary liquid

flows, which enabled rapid bubble collapse by direct condensation

in the main microchannel. This developed structure was shown to

be self-sustainable by creating strong mixing by collapsing bubble

slugs in the main channels at high frequency, but also as an effec-

tive method in reducing flow resistances and as artificial nucle-

ation sites.

It is a challenging task to generate appreciable mixing in micro-

channels without active control since flows in microchannels are

dominated by viscous and capillary flows [4,22,23], which tend to

suppress mixing. This, in turn, is detrimental to heat transfer at

themicro scale asmixing enhances advection heat transfer. In order

to create self-sustainable strongmixing inmicrochannels, an actua-

tion mechanism similar to water hammer, which can be enabled by

self-sustained flow control techniques, is desirable. In several previ-

ous studies, two-phase flow was actively enhanced by generating

0017-9310/$ - see front matter � 2012 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2012.11.057

⇑ Corresponding author. Tel.: +1 803 777 7155.

E-mail address: li01@mailbox.sc.edu (C. Li).

International Journal of Heat and Mass Transfer 58 (2013) 402–412

Contents lists available at SciVerse ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier .com/locate / i jhmt

seed bubbles [12–14]. Additionally, non-sustainable water hammer

pulses driven by bubble growth and collapse were observed and

modeled in a single microchannel [28]. However, self-sustainable

high frequency TPOs (pulses) powered by bubbles in microchannels

have not been studied in enhancing flow boiling before. In this

study, enhanced flow boiling by self-sustained high frequency TPOs

(pulses) powered by vapor bubble growth and collapse, which were

experimentally demonstrated in microchannels [29], was investi-

gated. Significant flow boiling heat transfer enhancement including

CHF values under constant heat fluxeswas achieved. Superior stabil-

ities in terms ofworking temperature andpressure drop during flow

boiling in microchannels were observed. Moreover, HTCs were

greatly enhanced with significant reduction in the pressure drop.

The working mechanisms of heat transfer enhancement under

investigation are discussed and revealed through visualization

study. Further optimization is proposed.

2. Experimental apparatus and procedures

2.1. Design of the microchannel architecture

The micro device consists of four main channels (H = 250 lm;

W = 200 lm; L = 10 mm) and each was connected to two auxiliary

channels (H = 250 lm; W = 50 lm; L = 5 mm) through a 20 lmopening (Fig. 1). The auxiliary channels were open to the inlet

manifold (Fig. 1e), but IRs (H = 250 lm; W = 20 lm; L = 400 lm)

were placed in the inlet to each main channels to trap bubbles. A

1 mm diameter inlet port, 1 mm diameter outlet port, and two

1 mm diameter pressure ports were fabricated in the micro heat

exchanger. The inlet to exit pressure drop was measured between

the two pressure ports. To minimize heat loss, two thermal isola-

tion gaps (air gaps) were etched on both sides of microchannel ar-

rays as shown in Fig. 1a. Flow stabilizers at the entrance to the inlet

manifold were formed to evenly distribute the flow. Additionally,

air gaps were etched between the auxiliary channels and the main

channels to reduce conjugate heating, and hence, to decouple ther-

mal interaction between channels. As shown in Fig. 1f, two open-

ings (nozzles) on both sidewalls of the main channel were made

to connect the two auxiliary channels with a cross junction located

at the middle of the axial position of the main channel. A micro

heater, which was made of a thin aluminum film, was deposited

onto the back side of silicon microchannels (Fig. 2). The heater area

(10 mm � 2 mm), which was identical to the total base area of

microchannel arrays and walls, served as a thermistor to measure

the average temperature of the heater.

2.2. Design and fabrication of the microchannel device

The microchannel heat exchangers were made from a silicon

wafer bonded to a Pyrex wafer by standard microfabrication pro-

cesses (Fig. 3). This process started with a double-side-polished

n-type h100i silicon wafer. First, 1 lm thick thermal oxide layers

were grown on both side of the silicon wafer. The silicon oxide film

provided electrical insulation for the micro heaters and acted as a

mask for deep reactive ion etching (DRIE) in subsequent microfab-

rication steps. An additional 500 nm thick oxide layer was then

deposited onto the front side as a shield layer by plasma-enhanced

chemical vapor deposition (PECVD). Next, a 7 nm thick adhesive

layer of titanium was deposited on the backside of the silicon wa-

fer by DC sputtering followed by a 1 lm thick layer of aluminum.

Once the thin films were successfully deposited, a thin film heater

was formed on the backside of the wafer by photolithography and

wet etching (Fig. 2b). A 500 nm thick PECVD oxide layer was then

deposited to protect the thin film heater in subsequent fabrication

processes.

After the heater was formed on the backside, a pattern mask of

microchannels on the top side of the wafer were etched on the sil-

icon oxide through photolithography and reactive ion etching

(RIE). The area under the oxide mask was protected and the

remaining areas were etched out to create 250 lm deep trenches

by DRIE. The DRIE process formed deep vertical sidewalls with a

root mean square (RMS) roughness of �300 nm [30]. The inlet port,

the outlet port, the air gaps, and the pressure ports were formed on

the backside of the wafer by DRIE. Photolithography and wet etch-

ing (6:1 buffered oxide etchant) was used to remove patterned

oxide in preparation for DRIE. Then through holes were etched

through the wafer by DRIE (Fig. 3d). RIE was used to remove pat-

terned oxide coatings on the backside to expose the contact pads.

Nomenclature

Ab base area, m2

Af fin surface, m2

As surface area, m2

At total heat transfer area, m2

Cp heat capacity at constant pressure, J/kg KG mass flux, kg/m2 s�h average heat transfer coefficient, W/m2 Khfg latent heat of vaporization, kJ/kgH channel height, mI electrical current, Ak thermal conductivity, W/m KL length, mm parameter for pin efficiency_m mass flow rate, kg/sN number of microchannelsp pressure, kPaP power, Wq00 heat flux, W/cm2

R electrical resistance, XSl slope of linear function, X/Kt thickness between heater and microchannel base, mT temperature, �C

T average temperature, �CV electrical voltage, VW microchannel width, m

Greek symbolsg efficiencyv vapor quality

Subscripts

a ambientb baseCHF critical heat fluxe exiteff effectivef finHTC heat transfer coefficienti inletl linears surfacesat saturatedsp single-phasetp two-phase

F. Yang et al. / International Journal of Heat and Mass Transfer 58 (2013) 402–412 403

A Pyrex glass wafer was anodically bonded to the silicon sub-

strate to seal the device as shown in Fig. 3e. The transparent glass

cover also served as a window for visualization studies. The indi-

vidual microchannel test chips (length 30 mm; width 10 mm;

thickness 1 mm) were cut from the wafer by a dice–saw.

2.3. Experimental setup

Fig. 4 depicts the two-phase apparatus used to conduct the

experiments. The microchannel test chip was placed at the

middle of the test package module (Fig. 4a), which provided

hydraulic ports and electrical connections. Mechanical fasten

units consist of two fasten bolts and two holding clamps helped

position and secure the microchannel test chips by mechanical

force. The microchannel test chips were supported at their two

ends and with the middle suspended. This packaging was care-

fully designed to minimize the stresses on the thin film therm-

istors/heater and to reduce heat loss. Six micro o-rings were

placed between the silicon microchannel test chip and the test

package to achieve mechanical seals. Probe-pins were placed un-

der the backside of the device and two bolts were used to adjust

the heights of the probe-pins to minimize electrical contact

resistance. Two pressure transducers were connected to the

pressure ports to measure pressure drops.

Fig. 1. (a) 3D CAD model of the present microchannel architecture, (b) and (c) magnified views of entrance and cross-junction structures, (d) and (e) top-viewed of cross-

junctions and entrances, (f) and (g) SEM images of the cross-junction.

404 F. Yang et al. / International Journal of Heat and Mass Transfer 58 (2013) 402–412

Major components of the experimental setup include an optical

imaging system, a data acquisition unit, and an open loop for cool-

ant supply (Fig. 4b). A pressurized water tank was used to supply

DI water, which was degassed prior to tests and pumped by com-

pressed nitrogen (N2). Mass fluxes were measured by a Sensiron

ASL1600 flowmeter with 0.03 kg m�2 s�1 resolution. Electrical

power was supplied by a high precision digital programmable

power supply. The voltage on the micro heater was measured by

an Agilent digital multimeter. Two thermocouples were used to

measure the inlet and outlet fluid temperatures. Flow rate, local

pressure, inlet and outlet temperature, and the heater’s voltage

and current were recorded automatically by a customized data

acquisition system developed from NI LabVIEW�. A visualization

system comprised of a high-speed camera (Phantom V 7.3) with

256 � 256 pixels at approximate 40,000 frames per second and

an Olympus microscope (BX-51) with 400 � amplifications was

developed to study the bubble dynamics.

2.4. Experimental procedure

Prior to experiments, the heat loss as a function of temperature

difference between the micro heat exchanger and the ambient was

evaluated. The temperature of the device in steady state was plot-

ted as a function of input heat fluxes without fluid flows. Thus, the

heat losses as a function of steady state temperature difference was

obtained by linear curve fit. The curve was then used to estimate

heat loss with high accuracy [31]. The heater, which also func-

tioned as a thermistor, was calibrated in an isothermal oven with

a proportional–integral–derivative (PID) controller. The tempera-

ture vs. electric resistance curve was generated using a linear curve

fit. The confidence of the correlation coefficient was estimated to

be higher than 0.9999.

After assembly of the microchannel device on the test package,

the flow rate was kept constant at a set value ranging from 150 to

480 kg/m2 s. Uniform heat flux was applied by a digital power sup-

ply through the heater at a step of approximate 2 W until

approaching CHF conditions. At each step, the data acquisition sys-

tem recorded 120 sets of steady state experimental data including

voltages, currents, local pressures and temperatures at inlet and

outlet at 4 min intervals.

3. Data reduction

3.1. Flow boiling heat transfer rate

The electrical input power and resistance of the heater was cal-

culated as, respectively,

P ¼ V � I ð1Þ

and

R ¼ V=I ð2Þ

Then heat loss, Ploss, between the environment and the testing chip

was deducted from P to calculate the effective power

Peff ¼ P � Ploss ð3Þ

The average temperature of the thermistor (i.e., the thin film heater)

was calculated as,

Theater ¼ SlðR� RaÞ ð4Þ

where Ra is the resistance of the micro heater at room temperature

and Sl is the slope of the heaters’ temperature vs. electrical resis-

tance curve. The average wall temperature of the base area of the

microchannel heat exchanger was estimated from the heater as

T ¼ Theater �q00eff t

ksð5Þ

where q00eff ¼ Peff =Ab.

The microchannels length (L) was divided into two sections:

single-phase region (Lsp) and two-phase region (Ltp) (Fig. 5). The

length of the two sections was observed to vary significantly with

exit vapor quality as shown in Fig. 5(b–d). The average two-phase

heat transfer coefficient, �htp, was used to evaluate flow boiling heat

transfer performance in the two-phase region (Ltp) by excluding

the weight of the single-phase heat transfer from the average tem-

perature. The single-phase heat transfer coefficient was calculated

as,

Fig. 2. Built-in micro heater (dimension in mm).

Fig. 3. Major microfabrication steps.

F. Yang et al. / International Journal of Heat and Mass Transfer 58 (2013) 402–412 405

�hsp ¼Peff

At � NAf ð1� gf Þh i

Tsp � ðT i þ TeÞ=2� �

ð6Þ

where the pin fin efficiency was estimated from gf = tanh(mH)/mH

and m ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2�hspðLþW f Þ=ksW f Lq

. The average temperature in the sin-

gle-phase region is the mean value of the inlet and exit surface tem-

peratures in this region.

Tsp ¼T in;sp þ Texit;sp

2ð7Þ

In the above equation, the inlet and the exit surface temperatures

were estimated as,

T in;sp ¼ T i þPeff

�hspAs

Texit;sp ¼ Tsat þPeff

�hspAs

ð8Þ

The average temperature of the two-phase heat transfer region, T tp,

was obtained by a weighted average method [32] in terms of the

single-phase and average wall temperatures:

T tp ¼TL� TspLsp

Ltpð9Þ

These length values (Lsp and Ltp) were measured through visualiza-

tion. Taking into account fin effects on a single microchannel, the

average two-phase heat transfer coefficient was calculated as,

Fig. 4. (a) An exploded 3D model of the test package module, which provide hydraulic and electrical connections, (b) the two-phase testing platform.

406 F. Yang et al. / International Journal of Heat and Mass Transfer 58 (2013) 402–412

�htp ¼Peff

P

WLþ 2HLgf

� �� �

T tp � Tsat

� �

ð10Þ

Because the thermal conductivity of Pyrex glass is approximately 1%

of silicon, the interface between the microchannel walls and the

cover glass was assumed to be thermally insulated in the fin

approximation. Then, �htp was iteratively obtained from Eqs. (1)–

(9). Additionally, the exit vapor quality was calculated with mass

flow rate and net input power according to:

v ¼Peff � _mCpðTsat � T iÞ

_mhfg

ð11Þ

3.2. Uncertainty analysis

Uncertainties of experimental variables are given in Table 1.

Uncertainties of measured values were adopted from the manufac-

turers’ specification sheets, and the uncertainties of derived vari-

ables were calculated according to the propagation of uncertainty

analysis [32].

4. Results and discussion

4.1. Flow boiling curve

The average wall temperature as a function of effective heat flux

at different mass fluxes is shown in Fig. 6. Bubbly flow was ob-

served following the ONB (Fig. 5a and b) during which the vapor

quality was relatively low and interactions between isolated bub-

bles were not significant. The average superheat at the ONB in this

study was approximately 2 �C, which was significantly lower than

the prediction by Liu et al. model (i.e., �12 �C) [33]. In the current

microchannel architecture, the nozzles/openings on the side walls

of the main channels (Fig. 1g) appear to function as artificial nucle-

ation sites as illustrated in Fig. 5a. These artificial nucleation cavi-

ties can reduce the superheat temperature at the ONB, similar to

observations in previous studies on engineered surfaces in micro-

channels [34]. Fig. 5a and b shows that bubbles were initially

nucleating near the nozzles on the side walls before propagating

in the posterior region of the main channels (Ltp). Onset of fully

developed boiling (OFB) was defined as the first occurrence of

nucleate boiling in all sections of the microchannels including

the main channels (Lsp section in Fig. 5a) and the auxiliary chan-

nels. When heat fluxes exceeded the OFB, explosive bubbles grew

inside the microchannels leading to direct condensation of vapor

slugs. Once vapor slugs directly contacted the sub-cooled liquid,

they started to shrink due to the direct condensation. These vapor

slugs completely collapsed in the microchannel when the conden-

sation rate was sufficiently high as shown in Fig. 5b. As a result, the

high frequency self-sustained two-phase oscillations were success-

fully created and a dramatic enhancement in flow boiling heat

transfer was observed. The detailed oscillation mechanism was

reported in previous study [29].

4.2. Two-phase heat transfer

Two-phase heat transfer coefficient as a function of effective

heat flux and vapor quality is depicted in Fig. 7. The heat transfer

coefficient initially increased with heat flux during fully developed

boiling and then gradually decreased.

The bubble ebullition cycle began in the vicinity of the nozzles

(highlighted by a dashed line in Fig. 5b). When the heat flux was

below the threshold required inducing explosive boiling, and

hence, activating two-phase oscillations, the single-phase region

(Lsp) and two-phase region (Ltp) had distinct boundaries (Fig. 5a).

With increasing heat flux, fully developed nucleate boiling gradu-

ally dominated in the region upstream of the nozzles and in the

auxiliary channels (Fig. 5c). The corresponding heat transfer coeffi-

cients were observed to increase considerably until reaching a

peak that varied for different mass fluxes. As the vapor quality in-

creased, the two-phase heat transfer coefficients reached their

maximum between exit qualities, v, between 0.15 and 0.3. The

maximum heat transfer coefficient achieved in this study was

approximately 120 kW/m2 K at G = 430 kg/m2 s and G = 480 kg/

m2 s. The mechanism responsible for this high heat transfer coeffi-

cient is believed to be related to the strong mixing induced by the

high frequency TPOs resulting from the vapor slug growth and

collapse process. The magnitude of the TPO frequency was

Fig. 5. (a) Single-phase flow and two-phase flow in the present microchannel

architecture. (b) Bubble growth at incipient boiling stage, v = 0.10. (c) Vapor slugs

in all sections of microchannel during fully developed boiling, v = 0.27. (d)

Persistent vapor slug was observed, which cannot be removed in the front section

of a main channel when boiling was approaching CHF, v = 0.49.

Table 1

Uncertainties of major parameters.

Name of variables Errors

Flow rate, Q 0.05%

Voltage on the heater, V 0.10%

Current on the heater, I 0.10%

Ambient temperature, Ta 0.1 �C

Electrical power on the heater, P 0.20%

Electrical resistance, R 0.20%

Average temperature, T 0.8 �C

Heat transfer coefficient, �h 2.20%

F. Yang et al. / International Journal of Heat and Mass Transfer 58 (2013) 402–412 407

approximately 100 Hz and the highest frequency in the present

microchannel architecture was over 600 Hz [29]. Heat transfer in

microchannels during flow boiling is greatly enhanced by the

strong mixing induced at such high frequencies. This is because

of the high frequency of liquid rewetting, mixing promoted advec-

tion, and evaporation, and the induced nucleation of bubbles

[29,35].

The gradual decline of the heat transfer coefficient is linked to

the formation of vapor slugs with exit qualities above a certain

threshold value in the main channels between the cross-junction

and the inlet (white areas in Fig. 5d) at high heat fluxes where

the high frequency TPOs have not affected the flow. The absence

of the high frequency TPOs mixing, forced convection and rewett-

ing caused nearly 50% of microchannel volume filled by vapor,

and hence, a dramatic decrease in the heat transfer coefficient.

The appearance of such a vapor slug can be a combined result of

a declined condensation rate and an increased evaporation rate

at high heat flux. Optimization of the location and number of

cross-junctions could improve this situation by increasing direct

condensation rates.

4.3. Comparisons between the present architecture and microchannels

with reentry cavities (RCs) and IRs

Comparisons among the current microchannel architecture and

those microchannels with RCs [32] and IRs [23,36] were made and

are shown in Figs. 8 and 9. In Fig. 8, three types of architectures

were compared to reveal the mechanism governing heat transfer

enhancement. These three different structures have identical

microchannel dimensions and IRs. The baseline microchannel

structure only has IRs and is denoted by IR architecture. The second

type of microchannel structure with IRs and RCs is termed RC archi-

tecture. Experimental data for RC architecture in Ref. [32] were se-

lected because they had the similar dimensions in the main

channels and IRs. Furthermore, all experimental data were reduced

by the same method. RCs have been demonstrated to enhance flow

boiling in microchannels [32,37,38] as shown in Fig. 8.

There are three primary approaches to enhance flow boiling

heat transfer: cavitations induced by IRs [36], artificial cavities

[32,37,38], and strong mixing. As illustrated in Fig. 8, the IR archi-

tecture had the lowest performance and the RC architecture is in

between the IR and the present microchannel structures. The RC

architecture in Ref. [32] performed better than the IR because it

had a much higher nucleation site density. However, the enhanced

the nucleation site density cannot explain the highest heat transfer

rate obtained on the present microchannel structure that has less

nucleation site density than the RC architecture does. The primary

mechanism behind the highest heat transfer rate should be the in-

duced mixing by the self-sustained high frequency TPOs.

At low vapor qualities, deviations in heat transfer coefficient be-

tween the present microchannel architecture and the RC architec-

ture developed by Kos�ar et al. [32] are insignificant as shown in

CHF

ONB

-50 0 50 100 150 200 250 300 350 400 450 500 550 600

20

40

60

80

100

120

140

160

180

200

Wal

l T

emp

erat

ure

()

Effective Heat Flux (W/cm2)

150 Kg/m2·s

250 Kg/m2·s

325 Kg/m2·s

380 Kg/m2·s

430 Kg/m2·s

480 Kg/m2·s

OFB

Fig. 6. Boiling curves: average wall temperature T vs. q00eff .

(a)

20

40

60

80

100

120

140

160 150 Kg/m2·s250 Kg/m2·s325 Kg/m2·s380 Kg/m2·s430 Kg/m2·s480 Kg/m2·s

Effective heat flux (W/cm2)

Hea

t T

ransf

er C

oef

fici

ent

(kW

/m2·K

)

50 100 150 200 250 300 350 400 450 500 550 600

0.0 0.1 0.2 0.3 0.4 0.5 0.620

40

60

80

100

120

140

160

Quality χ

150 Kg/m2·s250 Kg/m2·s325 Kg/m2·s380 Kg/m2·s430 Kg/m2·s480 Kg/m2·s

Hea

t T

ransf

er C

oef

fici

ent

(kW

/m2·K

) Fully

Developed

Boiling

(b)Initial

Boiling

Fig. 7. (a) Average two-phase heat transfer coefficient �htp vs. effective heat flux q00eff ,

(b) �htp vs. exit vapor quality v.

408 F. Yang et al. / International Journal of Heat and Mass Transfer 58 (2013) 402–412

Fig. 8a. Although the heat transfer coefficient was dramatically in-

creased during fully developed boiling, the heat transfer coefficient

in the present microchannels rapidly decline after reaching a max-

imum heat transfer coefficient and eventually overlapping with the

boiling curve reported in [32]. Such an enhancement is believed to

be a result of the induced mixing at low vapor qualities. The

enhancement of the RCs is not significant at low heat fluxes for

both designs. Furthermore, the heat transfer coefficient in the pres-

ent heat exchangers are lower prior to reaching fully developed

boiling than these reported in [32] for moderate mass fluxes

(Fig. 8b).

The number of RCs are much larger in Ref. [32] than in the cur-

rent microchannel architecture. Therefore, enhanced nucleation by

RCs in [32] resulted in a higher heat transfer coefficient compared

to the present architecture when nucleate boiling dominated, as

the present design consisted of only two artificial nucleation sites

at the center of main channels. As a result, its HTCs were close to

that of the IR architecture qualities corresponding to nucleate boil-

ing. However, heat transfer rate in the present microchannel struc-

ture is well above that reported in [32] for vapor qualities above

0.15, which suggests that the strong induced mixing inside the

microchannels plays a critical role in enhancing two-phase heat

transfer. In summary, heat transfer coefficient in the present

microchannel architecture was enhanced by up to 149% (Fig. 8a)

and 120% (Fig. 8b) with approximately identical mass fluxes and

40

50

60

70

80

90

100

110

250 Kg/m2·s; present325 Kg/m2·s; present302 Kg/m2·s; RC302 Kg/m2·s; IR

Hea

t T

ransf

er C

oef

.(k

W/m

2·K

)

(b)

Quality χ

Initial

Boiling

Fully Developed Boiling

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.720

30

40

50

60

70

150 Kg/m2·s;

present.

166 Kg/m2·s;

RC

166 Kg/m2·s;

IR

Hea

t T

ran

sfer

Co

ef.

(kW

/m2·K

)

(a)

Quality χ

Initial

Boiling

Fully Developed Boiling

Fig. 8. Comparisons of HTCs as functions of exit quality for three microchannel

structures: the present architecture, RC [32] and IR [23,36] architectures. (a) Low

mass flux, and (b) moderate mass fluxes.

0 200 400 600 800 1000 1200 1400 160010

100

1000

Pre

ssure

Dro

p (

kP

a)

Mass flux (Kg/m2·s)

150 W/cm2 Present150 W/cm2 IR250 W/cm2 Present250 W/cm2 IR

ONB

Reduced

∆p

Fig. 9. Dp-G curves of flow boiling in present microchannel architecture and IR

architectures.

0 50 100 150 200 250

133

134

135

136

137

138

Time (s)

Wal

l T

emp

erat

ure

()

80.0

80.5

81.0

81.5

82.0

82.5

83.0

83.5

84.0

84.5

Pressu

re Dro

p(k

Pa)

Wall Temperature

Pressure Drop

(b)

Bulk

Boiling

Fully

Developed

Boiling

Subcooled

Single-phase

Liquid Flow

Superheated

Single-phase

Vapor Flow

G

∆p

ONB

Unstable

Boiling

Constant

Heat Flux

a

bc

d

e

Negative

Slopef

(a)

OFI

Fig. 10. (a) Classic Dp-G curve of flow boiling in plain wall microchannels. (b)

Transient wall temperatures and pressure drops in 240 s at a mass flux of 380 kg/

m2 s and an effective heat flux of 296.6 W/cm2 in present microchannel

architecture.

F. Yang et al. / International Journal of Heat and Mass Transfer 58 (2013) 402–412 409

heat fluxes compared to the IR configuration and 87% and 57%

compared to the RC configurations developed in [32].

The pressure drop-mass flux (Dp-G) curves for heat fluxes of 150

and 250W/cm2 are shown in Fig. 9. A pressure drop reduction be-

tween 71% and 90% compared to IR configuration for mass flux

ranging from 400 to 1400 kg/m2 s were observed. The primary

objectives of the microchannel architecture were to create sustain-

able high frequency TPOs to promote flow boiling and to reduce

hydraulic resistance by managing vapor slug expansion rate and

by introducing bypasses. The significant reduction of the pressure

drop can be a result of the effective elimination of the confinements

of compressible vapor bubbles enabled by high frequency bubble

collapse as detailed in Section 4.5 as well as the increased cross-

sectional area from the auxiliary channels (Fig. 1f). For mass fluxes

below 400 kg/m2 s, the Dp-G curves of the two configurations

gradually converged because pressure drop of single-phase vapor

flow is dominated at high exit vapor quality.

Additionally, as illustrated in Fig. 9, the ONB in the present

microchannels were observed to be greatly reduced compared to

microchannels with smooth walls for a given heat flux. This is be-

cause the nozzles on both walls of the microchannels also serve as

artificial nucleation sites as discussed in Section 4.1.

4.4. Two-phase flow instabilities

The Dp-G curves in Fig. 10a are used to analyze the Ledinegg

instability in microchannels. At high mass fluxes and for fixed heat

flux, the pressure drop initially decreases with decreasing mass

flux (segment a–b in Fig. 10a) before reaching a minimum termed

the onset of flow instability (OFI). In microchannels, it has been

Fig. 11. High frequency two-phase oscillations powered by bubble growth/collapse processes at a heat flux of 100 W/cm2 for a mass flux of 400 kg/m2 s.

410 F. Yang et al. / International Journal of Heat and Mass Transfer 58 (2013) 402–412

argued that the OFI corresponds to a mass flux only slightly less

than the ONB. As the mass flux continue to decrease past the OFI,

the pressure drop-mass flux slope changes course and becomes

negative (Segment c–e in Fig. 10a). As discussed in Boure et al.

[1], when this happens, i.e., when

@ðDpÞ

@G6 0 ð12Þ

the system is susceptible to the Ledinegg instability, which is linked

to the parallel channel instability [5]. In addition, upstream com-

pressible volume instability can be significant at the micro scale.

Such flow instabilities are characterized by low frequencies and

large amplitudes of flow rate, pressure drops, and heat transfer

coefficients [1,3] and usually result in premature CHF. IRs or orifices

have been developed and demonstrated as an effective method to

mitigate two-phase flow instabilities [2,4,10,21,39] by reshaping

the Dp-G curve such that the curve is rendered positive. The Dp-G

curves of the present heat exchanger at two heat fluxes, q00eff ¼ 150

and 250 W/cm2 are compared to microchannels with IRs in Fig. 9.

The slopes of these two Dp-G curves from microchannels with IRs

are positive, which enables high CHF by suppressing flow oscilla-

tions [7], but, with a significant penalty in the form of elevated pres-

sure drop [7,22,23]. It has been observed that the slopes in the

current systems were greatly moderated (the slope of Dp-G curve

at 150 W/cm2 is even positive). As a result, two-phase instabilities

in terms of pressure drop were suppressed. High CHF was demon-

strated, i.e., 552 W/cm2 at a moderate mass flux of 480 kg/m2 s

(Fig. 6), with a significant pressure drop reduction compared to

microchannels with IRs. Transient pressure drops and wall temper-

atures at a mass flux of 380 kg/m2 s and effective heat flux of

296.6 W/cm2 are shown in Fig. 10b. At such a high heat flux, pres-

sure drop and wall temperature have been observed to fluctuate

within 1.2% (1 kPa) and 1.5% (2 �C), which verify the stability of

the system.

4.5. Visualization study

To obtain better fundamental knowledge of the processes gov-

erning the heat transfer enhancement discussed above, a visualiza-

tion study was conducted using a high-speed imagery system

(Phantom V7.3) attached to Olympus microscope (BX-51). Fig. 11

shows seven frames selected from sequential images in 8 ms incre-

ments at 10,000 frames per second to represent the entire bubble

ebullition processes near a cross-junction where two auxiliary

channels and a main-channel were connected. The interactions be-

tween vapor slugs in auxiliary channels and subcooled liquid in the

entrance of the auxiliary channels are illustrated in Fig. 12.

Fig. 11 shows how vapor slugs filling auxiliary channels A and B

triggered a bubble growth cycles (marked as time t = 0). A thin li-

quid film between the bubble slugs and the walls was observed.

The vapor slugs in the auxiliary channels only expanded toward

the inlet, i.e., toward the subcooled region of the main channel,

due to the restriction of the converging nozzles. The vapor slug

in auxiliary channel Bwas observed to shrink due to direct conden-

sation at 0.6 ms. The large vapor slug in the main channel collapsed

due to direct condensation, which was induced by the subcooled

flow through auxiliary channel B by fluid pulses at 1.4 ms. The

movement of two-phase interfaces, which were observed andmea-

sured by high speed camera, determines the superficial velocity of

the fluid. The velocity of the liquid jetting was evaluated to be

nearly five times higher than the average liquid velocity from the

inlet, i.e., accelerated. Alternatively, the vapor slug in auxiliary

channel A started to shrink at 2.8 ms and completely collapsed at

3.9 ms, which led to a jetting flow via a converging nozzle and fur-

ther shrinkage of the vapor slug in the main channel. This jetting

flow is similar to water hammer in microchannels [28], however,

it is self-sustainable. The measured frequency of these pulses in

the auxiliary channels and the main channel were approximately

100 and 200 Hz, respectively, at a heat flux of 100 W/cm2 with a

mass flux of 400 kg/m2-s. The highest oscillation frequency

achieved in the present microchannel architecture was over

613 Hz [29]. Such high frequency two-phase oscillations were

shown not to affect the system temperature and pressure drop.

The third fluid pulse during this observation period was formed

and accelerated into the cross-junction at 7.1 ms. A large vapor

slug in the main channel was observed to completely collapse or

had been removed at 8 ms, which significantly limits the impacts

of bubble expansion rate and confinements as well as the influence

of capillary force on two-phase flows in the microchannel. The

walls in the microchannels can be wetted by high frequency

two-phase oscillations, which, consequently, enhance heat transfer

because of the thin film evaporation and convection. It should be

noted that nucleate boiling heat transfer was also greatly enhanced

due to induced cavitations during the collapse of large vapor slugs

[29].

Flow reversal driven by the rapid vapor slug expansion in the

auxiliary channels was observed. The complete process of a vapor

slug expansion and collapse due to direct condensation is shown in

Fig. 12. Flow reversal appeared at time 0 ms. The vapor slug ex-

panded toward the subcooled fluid in the inlet region due to the

restriction of the nozzles. The volume of vapor slug reached a max-

imum value at 2.1 ms when the evaporation and direct condensa-

tion heat transfer was approaching equilibrium. Disappearance of

the vapor slug fronts and the reassumed subcooled flow in the aux-

iliary channels indicated the collapse of vapor slug as well as the

completion of the vapor slug ebullition process.

Fig. 12. Vapor slug expansion and collapse in the entrance area. The working

condition is at a heat flux of 150 W/cm2 for a mass flux of 400 kg/m2 s.

F. Yang et al. / International Journal of Heat and Mass Transfer 58 (2013) 402–412 411

5. Conclusions

In this study, a new type of microchannel architecture was

developed and built to enhance flow boiling in microchannels

through self-sustained TPOs. Experiments demonstrated that this

microchannel architecture can considerably enhance flow boiling

heat transfer rate, achieve high flow boiling stabilities in terms of

working temperature and pressure drop, and enhance CHF with a

significant reduction in the pressure drop compared to microchan-

nels with RCs and IRs. A visualization study was conducted to

better understand the enhancement mechanisms. The main con-

clusions are summarized below:

� High frequency bubble growth/collapse process in microchan-

nels can be passively excited and sustained to create two-phase

flow oscillations and hence to generate strong mixing. This, in

turn, is an effective methodology to enhance flow boiling heat

transfer by promoting thin film evaporation, nucleate boiling,

and advection.

� The pressure drop can be significantly reduced through effective

management of confinements of compressible vapor slugs and

the high frequency collapse of vapor slugs.

� Flow boiling was observed to be stable in terms of working tem-

perature and pressure drop.

� Optimized configuration of the present microchannel architec-

ture by improving the direct condensation capabilities will fur-

ther improve flow boiling in microchannels.

� Since no extra activation method or moving parts were intro-

duced, the present microchannel architecture can achieve com-

pact and efficient cooling systems at lower cost and with higher

reliabilities compared with active approaches.

Acknowledgments

This work was supported by the startup funding of University of

South Carolina and the Office of Naval Research (Program Officer

Dr. Mark Spector) under Grant No. N000141210724. SEM figures

in this study were taken in Electron Microscopy Center at Univer-

sity of South Carolina. A part of microfabrication was performed at

the Cornell NanoScale Facility (CNF), a member of the National

Nanotechnology Infrastructure Network of United States, which

is supported by the National Science Foundation (Grant ECS-

0335765).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in

the online version, at http://dx.doi.org/10.1016/j.ijheatmass

transfer.2012.11.057.

References

[1] J.A. Boure, A.E. Bergles, L.S. Tong, Review of two-phase flow instability, Nucl.Eng. Des. 25 (2) (1973) 165–192.

[2] H.Y. Wu, P. Cheng, Boiling instability in parallel silicon microchannels atdifferent heat flux, Int. J. Heat Mass Transfer 47 (17–18) (2004) 3631–3641.

[3] A.E. Bergles, S.G. Kandlikar, On the nature of critical heat flux in microchannels,J. Heat Transfer-Trans. ASME 127 (1) (2005) 101–107.

[4] K.H. Chang, C. Pan, Two-phase flow instability for boiling in a microchannelheat sink, Int. J. Heat Mass Transfer 50 (11–12) (2007) 2078–2088.

[5] Y. Peles, Contemporary Perspectives on Flow Instabilities in Microchannels,Begell House Inc., 2012.

[6] G. Wang, P. Cheng, H. Wu, Unstable and stable flow boiling in parallelmicrochannels and in a single microchannel, Int. J. Heat Mass Transfer 50 (21–22) (2007) 4297–4310.

[7] G. Wang, P. Cheng, A.E. Bergles, Effects of inlet/outlet configurations on flowboiling instability in parallel microchannels, Int. J. Heat Mass Transfer 51 (9–10) (2008) 2267–2281.

[8] P. Balasubramanian, S.G. Kandlikar, Experimental study of flow patterns,pressure drop, and flow instabilities in parallel rectangular minichannels, HeatTransfer Eng. 26 (3) (2005) 20–27.

[9] J. Xu, J. Zhou, Y. Gan, Static and dynamic flow instability of a parallelmicrochannel heat sink at high heat fluxes, Energy Convers. Manage. 46 (2)(2005) 313–334.

[10] H.Y. Wu, P. Cheng, Visualization and measurements of periodic boiling insilicon microchannels, Int. J. Heat Mass Transfer 46 (14) (2003) 2603–2614.

[11] W.L. Qu, I. Mudawar, Measurement and correlation of critical heat flux in two-phase micro-channel heat sinks, Int. J. Heat Mass Transfer 47 (10–11) (2004)2045–2059.

[12] J. Xu, G. Liu, W. Zhang, Q. Li, B. Wang, Seed bubbles stabilize flow and heattransfer in parallel microchannels, Int. J. Multiph. Flow 35 (8) (2009) 773–790.

[13] G. Liu, J. Xu, Y. Yang, Seed bubbles trigger boiling heat transfer in siliconmicrochannels, Microfluid. Nanofluid. 8 (3) (2010) 341–359.

[14] G.H. Liu, J.L. Xu, Y.P. Yang, A.W. Zhang, Active control of flow and heat transferin silicon microchannels, J. Micromech. Microeng. 20 (4) (2010).

[15] P.-H. Chen, W.-C. Chen, S.H. Chang, Bubble growth and ink ejection process of athermal ink jet printhead, Int. J. Mech. Sci. 39 (6) (1997) 683–695.

[16] P. Deng, Y.-K. Lee, P. Cheng, The growth and collapse of a micro-bubble underpulse heating, Int. J. Heat Mass Transfer 46 (21) (2003) 4041–4050.

[17] H. Andersson, W. van der Wijngaart, P. Nilsson, P. Enoksson, G. Stemme, Avalve-less diffuser micropump for microfluidic analytical systems, Sens.Actuat. B-Chem. 72 (3) (2001) 259–265.

[18] S.G. Kandlikar, History, advances, and challenges in liquid flow and flowboiling heat transfer in microchannels: a critical review, J. Heat Transfer-Trans.ASME 134 (3) (2012).

[19] P. Cheng, G.D. Wang, X.J. Quan, Recent work on boiling and condensation inmicrochannels, J. Heat Transfer-Trans. ASME 131 (4) (2009).

[20] J.R. Thome, State-of-the-art overview of boiling and two-phase flows inmicrochannels, Heat Transfer Eng. 27 (9) (2006) 4–19.

[21] W.L. Qu, I. Mudawar, Measurement and prediction of pressure drop in two-phase micro-channel heat sinks, Int. J. Heat Mass Transfer 46 (15) (2003)2737–2753.

[22] S.G. Kandlikar, W.K. Kuan, D.A. Willistein, J. Borrelli, Stabilization of flowboiling in microchannels using pressure drop elements and fabricatednucleation sites, J. Heat Transfer-Trans. ASME 128 (4) (2006) 389–396.

[23] A. Kosar, C.J. Kuo, Y. Peles, Suppression of boiling flow oscillations in parallelmicrochannels by inlet restrictors, J. Heat Transfer-Trans. ASME 128 (3) (2006)251–260.

[24] I. Mudawar, Assessment of high-heat-flux thermal management schemes, IEEETrans. Compon. Packag. Technol. 24 (2) (2001) 122–141.

[25] Z.H. Liu, T.F. Tong, Y.H. Qiu, Critical heat flux of steady boiling for subcooledwater jet impingement on the flat stagnation zone, J. Heat Transfer-Trans.ASME 126 (2) (2004) 179–183.

[26] A. Bar-Cohen, M. Arik, M. Ohadi, Direct liquid cooling of high flux micro andnano electronic components, Proc. IEEE 94 (8) (2006) 1549–1570.

[27] S.G. Kandlikar, A.V. Bapat, Evaluation of jet impingement, spray andmicrochannel chip cooling options for high heat flux removal, Heat TransferEng. 28 (11) (2007) 911–923.

[28] S.G. Kandlikar, Scale effects on flow boiling heat transfer in microchannels: afundamental perspective, Int. J. Therm. Sci. 49 (7) (2010) 1073–1085.

[29] F. Yang, X. Dai, C. Li, High frequency microbubble-switched oscillationsmodulated by microfluidic transistors, Appl. Phys. Lett. 101 (7) (2012).073509-073504.

[30] D.W. Fogg, K.E. Goodson, Bubble-induced water hammer and cavitation inmicrochannel flow boiling, J. Heat Transfer-Trans. ASME 131 (12) (2009).

[31] S. Vengallatore, Y. Peles, L.R. Arana, S.M. Spearing, Self-assembly of micro- andnanoparticles on internal micromachined silicon surfaces, Sens. Actuat., A 113(1) (2004) 124–131.

[32] A. Kosar, C.J. Kuo, Y. Peles, Boiling heat transfer in rectangular microchannelswith reentrant cavities, Int. J. Heat Mass Transfer 48 (23–24) (2005) 4867–4886.

[33] S.J. Kline, F.A. McClintock, Describing uncertainties in single-sampleexperiments, Mech. Eng. 75 (1) (1953) 3–8.

[34] D. Liu, P.S. Lee, S.V. Garimella, Prediction of the onset of nucleate boiling inmicrochannel flow, Int. J. Heat Mass Transfer 48 (25–26) (2005) 5134–5149.

[35] D.W. Fogg, K.E. Goodson, Bubble-induced water hammer and cavitation inmicrochannel flow boiling, J. Heat Transfer 131 (12) (2009) 121006–121012.

[36] B. Schneider, A. Kosar, C.J. Kuo, C. Mishra, G.S. Cole, R.P. Scaringe, Y. Peles,Cavitation enhanced heat transfer in microchannels, J. Heat Transfer-Trans.ASME 128 (12) (2006) 1293–1301.

[37] H.T.H. Kubo, Hiroshi Honda, Effects of size and number density of micro-reentrant cavities on boiling heat transfer from a silicon chip immersed indegassed and gas-dissolved FC-72, J. Enhanc. Heat Transfer 6 (2–4) (1999)151–160.

[38] C.J. Kuo, Y. Peles, Flow boiling instabilities in microchannels and means formitigation by reentrant cavities, J. Heat Transfer-Trans. ASME 130 (7) (2008).

[39] J.E. Kennedy, G.M. Roach, M.F. Dowling, S.I. Abdel-Khalik, S.M. Ghiaasiaan, S.M.Jeter, Z.H. Quershi, The onset of flow instability in uniformly heated horizontalmicrochannels, J. Heat Transfer-Trans. ASME 122 (1) (2000) 118–125.

412 F. Yang et al. / International Journal of Heat and Mass Transfer 58 (2013) 402–412