Saturated flow boiling heat transfer and associated bubble characteristics of FC-72 on a heated micro-pin-finned silicon chip

(1)

Saturated ﬂow boiling heat transfer and associated

bubble characteristics of FC-72 on a heated

micro-pin-ﬁnned silicon chip

Y.M. Lie

a

, J.H. Ke

a

, W.R. Chang

c

, T.C. Cheng

b

, T.F. Lin

a,*

a

Department of Mechanical Engineering, National Chaio Tung University, Hsinchu, Taiwan, ROC b

National Nano Device Laboratories, Industrial Technology Research Institute, Hsinchu, Taiwan, ROC c

Energy and Resources Laboratories, Industrial Technology Research Institute, Hsinchu, Taiwan, ROC Received 22 December 2005; received in revised form 3 February 2007

Available online 11 April 2007

Abstract

An experiment is carried out here to investigate flow boiling heat transfer and associated bubble characteristics of FC-72 on a heated micro-pin-finned silicon chip flush-mounted in the bottom of a horizontal rectangular channel. Besides, three different micro-structures of the chip surface are examined, namely, the smooth, pin-finned 200 and pin-finned 100 surfaces. The pin-finned 200 and 100 surfaces, respectively, contain micro-pin-fins of size 200 lm 200 lm 70 lm (width length height) and 100 lm 100 lm 70 lm. The pitch of the fins is equal to the fin width for both surfaces. The effects of the FC-72 mass flux, imposed heat flux, and surface micro-structures of the silicon chip on the FC-72 saturated flow boiling characteristics are examined in detail. The experimental data show that an increase in the FC-72 mass flux causes a delay in the boiling incipience. However, the flow boiling heat transfer coefficient is not affected by the coolant mass flux. But adding the micro-pin-fin structures to the chip surfaces can effectively enhance the single-phase convection and flow boiling heat transfer. Moreover, the mean bubble departure diameter and active nucleation site density are reduced for a rise in the FC-72 mass flux. A higher coolant mass flux results in a higher mean bubble departure frequency. Furthermore, larger bubble departure diameter, higher bubble departure frequency, and higher active nucleation site density are observed at a higher imposed heat flux. We also note that adding the micro-pin-fins to the chips decrease the bubble departure diameter and increase the bubble depar-ture frequency. However, the departing bubbles are larger for the pin-finned 100 surface than the pin-finned 200 surface but the bubble departure frequency exhibits an opposite trend. Finally, empirical equations to correlate the present data for the FC-72 single-phase liquid convection and saturated flow boiling heat transfer coefficients and for the bubble characteristics are provided.

1. Introduction

The recent quick development of the IC (Integrated Cir-cuits) technology significantly reduces the size of the elec-tronic equipments and substantially increases the density of the power dissipation in the equipments. It is well known that the heat removal methods based on gas cooling are not sufficiently effective for the high-emission heat components.

The direct liquid cooling is better. Moreover, flow boiling is one of the most effective methods because the high latent heat involved in the process. The dielectric coolant FC-72 is appropriate for the electronics cooling. Up to the pres-ent, the bubble characteristics associated with the cooling of electronic equipments by flow boiling of dielectric liq-uids remain poorly understood. The situation is even worse for flow boiling on some enhanced surfaces. Hence the rela-tion between the bubble behaviors and heat transfer perfor-mance can not be delineated.

In the following the literature relevant to the present

study is reviewed. Mudawar and his colleagues[1,2]

doi:10.1016/j.ijheatmasstransfer.2007.02.010 *

Corresponding author. Tel.: +886 35 712121 55118; fax: +886 35 726440.

E-mail address:tﬂ[email protected](T.F. Lin).

(2)

flush-mounted heat sources for the flow velocity ranging from 13 to 400 cm/s. They found that the heat transfer coefficient in the single-phase region varied linearly with the flow velocity and the boiling incipience was delayed for a higher flow velocity. Similar experiments were

con-ducted by Heindel et al.[3]also for FC-72 and they noted

that at increasing velocity the heat ﬂux increased but the temperature overshoot at onset of nucleate boiling reduced. In the fully developed boiling, the velocity exhib-ited little eﬀect on the boiling curves, which was also

sup-ported by McGills et al. [4]. On the other hand, a higher

liquid subcooling resulted in a smaller temperature over-shoot at ONB. In vertical channel ﬂow boiling experiments

for FC-72, Tso et al.[5]found that the temperature of the

chip surface decreased with the increases in the flow veloc-ity and liquid subcooling in the partial boiling region. While in the fully-developed boiling region both the flow velocity and subcooling temperature had little effects on the chip surface temperature. The temperature of the chip surface for the flush-mounted chips was lower than the pro-truded chips in the partial boiling region, but they were nearly equal in the fully developed boiling region. Samant

and Simon [6] analyzed the heat transfer from a small

region to refrigerants R-113 and FC-72 and noted that as the ﬂow velocity and subcooling temperature increased,

the temperature excursion and boiling hysteresis appeared to decrease.

Ma and Chung [7] investigated bubble dynamics in

reduced gravity flow boiling of FC-72 over a thin gold film semi-transparent heater. The bubble departure size was found to be bigger in the micro-gravity than environment. At increasing flow rate, the bubble departure time and departure size decreased. This was also observed by Situ

et al. [8] later. Besides, they also noted that the bubble

growth rate dropped sharply after lift-oﬀ. In addition,

Yin et al. [9] examined the subcooled ﬂow boiling of

R-134a in an annular duct and found that both the bubble departure size and frequency reduced at increasing liquid

subcooling. Experiments conducted by Bang et al. [10]

and Chang et al. [11] for R-134a and water focused on

the behavior of near-wall bubbles in subcooled flow boil-ing. They identified four different two-phase flow patterns including the discrete attached bubbles, sliding bubbles, small coalesced bubbles and large coalesced bubbles or vapor clots at increasing heat flux. For the higher mass flux of the flow, the coalesced bubbles were smaller. Bang et al.

[12] further noticed the presence of the R-134a vapor

remants below the discrete bubbles and coalesced bubbles, and the presence of an interleaved liquid layer between the vapor remants and bubbles. Besides, the bubble layer could Nomenclature

Achip surface area of a bare chip (m2)

Af surface area of a single ﬁn (m2)

Bo Boiling number, Bo¼_Giq_fg, dimensionless

Bf ﬁn height (m)

cp speciﬁc heat (J/kg°C)

d_p; d_p mean dimensional and dimensionless bubble

departure diameter (m)

Dh hydraulic diameter of rectangular-channel (m)

E enhancement factor

f ; f mean dimensional and dimensionless bubble

departure frequency (s1)

Fsp, Fd,sat pin-fin factors for the effects of fin geometries,

dimensionless

Frl Froude number, Frl¼ G

2

q2

lgDh, dimensionless

g acceleration due to gravity (m/s2)

G mass ﬂux (kg/m2s)

hl, hr single and two-phase heat transfer coeﬃcients

(W/m2K)

H height (m)

ilv enthalpy of vaporization (J/kg K)

I measured current from DC power supply (A)

k thermal conductivity (W/m K)

L length (m)

N number of micro-pin-ﬁns

Nac mean active nucleation site density (n/m2)

Nconf Conﬁnement number, Nconf¼ðr=gDqÞ

0:5

Dh ,

dimen-sionless

Nul Nusselt number for liquid convection, Nul¼hlkLl,

dimensionless

P system pressure (kPa)

Pr liquid Prandtl number, Pr¼llcp

kl , dimensionless

Qe, Qt eﬀective and total heat transfer rates (W)

q00 _{average eﬀective imposed heat ﬂux (W/m}2

)

Rc,h resistances of the cooper and mica plates

Rel liquid Reynolds number, Rel¼GD_lh

l ,

dimension-less

Sf ﬁn space between two adjacent ﬁns (m)

Tchip; T0chip temperatures at the upper and lower surfaces

of chip (K)

Theater temperature of the heater surface (K)

Tin liquid temperature at test section inlet (K)

Tsat saturated temperature of FC-72 (K)

V measured voltage from DC power supply (V)

Vb volume of a mean departing volume (m3)

Wf width of a ﬁn (m)

Greek symbols

DTc,h temperature diﬀerenceðTheater T0chipÞ (K)

DTsat wall superheat, (Tchip Tsat)

Dq density diﬀerence (ql qv) (kg/m3)

e relative heat loss, dimensionless

ll dynamic viscosity (N s/m2)

ql, qv liquid and vapor densities (kg/m3)

(3)

be divided into two types, a near-wall bubble layer domi-nated by small bubbles and a following bubble layer pre-vailed by large coalesced bubbles. Using digital imaging

and analyzing techniques, Maurus et al. [13] examined

the bubble characteristics and local void fraction in subco-oling flow boiling. The bubble population was noted to increase with the heat flux and the bubble density reduced drastically at increasing mass flux. Besides, the bubble size increased at increasing heat flux and decreasing mass flux.

In a continuing study [14] they further showed that the

effects of the heat flux and mass flux on the bubble size dis-tribution were weak for small bubbles but became more pronounced for bigger bubbles. The total bubble life time, the remaining lifetime after the detachment process and the waiting time between two bubble cycles decreased signifi-cantly as the mass flux increased. Moreover, they pointed out that the bubble behavior was dominated by the temper-ature of the thermal boundary layer and the turbulence intensity due to the heat transport in the liquid near the interface. On the other hand, in vertical upflow and

down-ﬂow boiling Thorncroft et al. [15] found that the bubble

growth rate and departure diameter increased with the wall superheat but decreased with increasing mass ﬂux.

Flow boiling of FC-72 over microstud, microgroove, and cylindrical micro pin-ﬁn enhanced surfaces ﬂush-mounted on a vertical rectangular-channel wall was

exam-ined by Maddox and Mudawar[16,17]. The surface

micro-structures were found to signiﬁcantly enhance the heat transfer performance and reduce the boiling hysteresis. Heat transfer in pool boiling of FC-72 on silicon chips with the surface micro-structures of micro-pin-ﬁns was recently

investigated by Honda et al.[18,19]. They found that both

the nucleate boiling heat transfer and the critical heat flux were effectively enhanced by the micro-pin-fins. The boiling phenomena observed revealed that a small amount of vapor was left within the gap between the pin-fins when a growing bubble left the surface. On the other hand, Honda

and Wei[20]conducted a critical review on the boiling heat

transfer enhanced by the surface-structures. They indicated that all the surface structures including the micro-roughness, micro-reentrant, and micro-porous structures were helpful in reducing the boiling incipience superheat. Surface cavities were effective in increasing critical heat flux and in enhancing nucleate boiling. Generally, the micro-pin-fins were most effective in augmenting CHF and micro-porous structures were most effective in enhancing nucleate boiling heat transfer. Recently, Ramaswamy

et al. [21] examined the eﬀects of varying geometrical

parameters on boiling from microfabricated enhanced structures. They indicated that increasing pore size caused higher heat dissipation and the pore pitch had more signif-icant eﬀect on the heat transfer performance. Finally, heat transfer from the microporous ﬁnned surface was found to

have better performance than the plain ﬁnned surface[22].

The above literature review clearly indicates that the pool boiling heat transfer from micro-structured surface and the associated bubble dynamics have received some

attention. But heat transfer and bubble characteristics in flow boiling of dielectric coolants on micro-structured sur-faces remain largely unexplored. In this study an experi-ment is carried out to investigate the FC-72 flow boiling on a single silicon chip with micro-pin-fins on its surface. The chip is flush-mounted on the bottom of a horizontal rectangular channel.

2. Experimental apparatus and procedures

The experimental system established in the present

study, schematically depicted inFig. 1, consists of a

degas-sing unit, a coolant loop, a hot-water loop, and a cold-water loop. They are described in the following.

2.1. Degassing unit

Because air or non-condensable gas dissolved in the coolant FC-72 can signiﬁcantly aﬀect the heat transfer per-formance and nucleate boiling phenomena, we must degas the coolant before beginning the experiments. The degas-sing unit is a 8-liter tank with an electric-heated patch in it to heat the coolant to its boiling point and the coolant begins to evaporate. The air and non-condensable gas and a little amount of the coolant vapor can be removed from the released valve on the top of the tank. Besides, a pressure transducer and a thermocouple are equipped in the tank to measure the pressure and temperature of FC-72, respectively.

2.2. Coolant loop

After degassing FC-72, we can remove the non-condens-able gases possibly existing in the coolant loop by using a vacuum pump and then fill FC-72 into the loop. The cool-ant loop consists of a variable-speed magnetic micro-pump, a filter, a volume flow meter, a pre-heater, a test section including the inlet and outlet sections, a condenser, and a receiver. The coolant flow rate is mainly controlled by an AC motor and can be further adjusted by regulating a bypass valve. The coolant FC-72 at the outlet of the mag-netic micro-pump must be kept subcooled to avoid any vapor flow through the volume flow meter. Finally, a vapor–liquid mixture is generated in the test section when the coolant moves over the heated silicon chip. The vapor flow leaving the test section is re-liquefied by the condenser in the cold-water loop. After leaving the condenser, the liquid FC-72 flows back to the receiver. An accumulator is connected to a high-pressure nitrogen tank to dampen the fluctuations of the coolant flow rate and pressure. The filter is used to filter the impurities and non-condens-able gas possibly existing in the loop. Varying the temper-ature and flow rate of the hot-water flowing through the pre-heater allows us to control the pressure of the coolant loop. Two absolute pressure transducers are installed at the inlet and outlet of the test section with a resolution up to ±2 kPa. All the coolant and water temperatures are

(4)

mea-sured by copper-constantan thermocouples (T-type) with a

calibrated accuracy of ±0.2°C.

2.3. Test section

The test section mainly contains a silicon chip flush-mounted on the channel bottom. The flow-channel consists of a gradually converging inlet section, the main test sec-tion, and a gradually diverging exit section. They are all made of stainless steel plate. The installation of the inlet and exit sections avoids the sudden change in the cross sec-tion of the channel. The test secsec-tion is a rectangular chan-nel of 20 mm in width, 5 mm in height, and 150 mm in length. The chip is mounted around the geometric center of the bottom plate of the test section. An observational window is installed on the upper lid of the test section right above the chip. The temperature and pressure of the FC-72 flow at the inlet and exit of the test section are measured by calibrated thermocouples and pressure transducers. The

sil-icon chip module schematically shown inFig. 2includes a

hollow cylindrical Teflon block, a cylindrical Teflon bolt, a silicon chip, a copper plate, two pieces of mica, a Teflon plate, and an electric-heater. To reduce the thermal contact resistance, thermal conducting grease is filled into the gaps between all the adjacent plates. The surface area of the

sil-icon chip is 10 mm 10 mm and the chip is heated by

pass-ing DC current through the electric-heater. Besides, three thermocouples are fixed at the back surface of the silicon chip to estimate the surface temperature of the silicon chip and another thermocouple is fixed at the upper surface of the electric-heater to measure its surface temperature. The square micro-pin-fins on the silicon chip are fabricated by the semiconductor manufacturing technique. Two

sur-face micro-structures in the form of square micro-pin-ﬁns are manufactured on the silicon chips and each individual

ﬁn has the same size of 200 lm 200 lm 70 lm for the

pin-ﬁnned 200 surface and 100 lm 100 lm 70 lm for

the pin-finned 100 surface. The space between the two adja-cent fins (the fin pitch) is about equal to the fin width and the detailed photographs of the arrays of selected

micro-FILTER / DRYER HOT-WATER THERMOSTAT FLOWMETER N2 ACCUMULATOR PREHEATER FLOWMETER T T P T TEST SECTION DC POWER SUPPLY MICRO PUMP BYPASS VALVE COOLANT 8 LITER RECEIVER _COLD-WATER THERMOSTAT DEGASSING UNIT T P CONDENSER T FLOWMETER P T P

Fig. 1. Schematic diagram of experimental apparatus.

Silicon chip Electric-heater Mica Mica Copper Teflon Stainless rectangular flow-channel Cylindrical-hollow Teflon block Cylindrical Teflon bolt Screw holes * 4 Top View

A - A Cross-sectional View

A A

(5)

pin-ﬁns on the chips taken by the electronic microscope are

shown inFig. 3.

2.4. Hot-water loop

In order to maintain the coolant FC-72 at the preset temperature at the test section inlet, a hot-water loop is used to preheat the coolant before it arrives at the test sec-tion inlet. The hot-water loop includes a thermostat with a 20-liter hot-water container, a 2-kW heater and a 0.5-hp water pump which can drive the water at a specified flow rate to the pre-heater. Besides, a bypass valve in the loop can further adjust the water flow rate. The connecting pipe between the pre-heater and test section is thermally insu-lated with a 5-cm thick polyethylene layer.

2.5. Cold-water loop

The cold-water loop is designed for condensing the liquid–vapor mixture of FC-72 from the test section. The

maximum cooling capacity of the thermostat is

2000 Kcal/h. The cold water at a specific flow rate is driven by a 0.5-hp pump to the condenser and a bypass loop is provided to adjust the flow rate. By adjusting the tempera-ture and flow rate of the cold water, the bulk temperatempera-ture of FC-72 in the condenser can be controlled at a preset level.

2.6. DC power supply

A 30V-3A DC power supply delivers the required elec-tric current to the elecelec-tric heater. A Yokogawa data logger is used to measure the DC voltage across and DC current passing through the heater with an accuracy of ±1%. Thus the power input to the heater can be calculated.

2.7. Data acquisition

The data acquisition system employed to acquire and process the data from various transducers is a 20-channel Fig. 3. Photographs of micro-pin-ﬁns on the silicon chips taken by SEM.

(6)

data logger (YOKOGAWA DA-100) along with a personal computer. The voltage signals from the thermocouples, pressure transducers, and volume ﬂow-meters are con-verted to the temperature, pressure, and volume ﬂow rate by the internal calibration equations in the computer and are displayed on the screen simultaneously.

2.8. Optical measurement technique

The photographic apparatus consists of a high speed digital video camera (IDT High-speed CMOS Digital Camera), a micro-lens (Optem Zoom 160), a three-dimen-sional positioning mechanism, and a personal computer. The high-speed camera can take photographs up to 143,307 frames/s. Here, a recording rate of 5000 frames/s is adopted to obtain the images of the bubble ebullition processes. The positioning mechanism is used to hold the camera at the required accurate position. The data for the bubble characteristics are collected in the regions near the geometric center of the chip surface. After the flow reaches a statistically steady state, we start recording the boiling activity. The high speed camera stores the images which are later downloaded to a personal computer. Then, the mean bubble departure diameter and frequency and active nucleation site density are calculated by viewing more than 500 frames for each case. In order to achieve the highest possible resolution and to eliminate errors in calibration, the camera lens is fixed at a constant focal length, resulting in a fixed viewing area. Typically, a total of over 150 bubble diameter measurements is used to con-struct the present data. The bubble departure frequency is measured by counting the total number of bubbles that emerge from the targeted heating surface area during a per-iod of a second.

2.9. Experimental procedures

Before conducting the experiments, the liquid FC-72 is degassed and then filled into the coolant receiver. Besides, the non-condensable gases in the coolant loop is evacuated. In each test, we first turn on the pump controller and set the inverter frequency to the required rotation rate of the AC motor to regulate the FC-72 flow rate to a preset level. Then the temperature and flow rate of the hot-water loop are selected so that the FC-72 temperature at the test sec-tion inlet can be maintained at a preset level. The imposed heat flux to the coolant in the test section is adjusted by varying the electric current delivered from the DC power supply. In addition, the current delivered to and voltage drop across the heater are measured. Temperature and flow rate of the cold water in the cold-water loop can be adjusted to condense the liquid–vapor mixture of FC-72 from the test section. Next, we regulate the FC-72 pressure at the test section inlet by adjusting the gate valve locating right after the test section exit. Meanwhile we use the bypass valve to further adjust the coolant flow rate to the required level. All measurements proceed when the

experi-mental system has reached statistically stable state. Finally, all the data channels are scanned every 1 s for a period of 30 s.

3. Data reduction

At ﬁrst, the total heat loss of the test section is evaluated

from the diﬀerence between the total power input Qtto the

silicon chip and the eﬀective power input Qeto the coolant

ﬂow over the chip. The total power input can be calculated from the measured voltage drop across and electric current

passing through the electric-heater, Qt= V I. The eﬀective

power input from the chip to the coolant is directly evalu-ated by assuming the 1-D heat conduction across the cop-per and mica plates sandwiched between the silicon chip and electric heater by neglecting the heat loss from the side-walls of the cooper and mica plates,

Q_e¼DTc;h

Rc;h

ð1Þ

where DTc;h¼ Theater T0chip is the temperature diﬀerence

between the upper surface of the heater and lower surface

of the chip. And Rc,h is the total thermal resistance from

the upper heater surface to the lower chip surface including

the resistances of the copper and mica plates. Here T0_chipis

the average measured temperature at the thermocouple locations on the lower surface of the chip. The imposed heat ﬂux at the chip surface is deﬁned as

q00¼ Qe=Achip ð2Þ

where Achipis the surface area of the bare chip. The relative

heat loss from the test section is deﬁned as

e¼ ðQt QeÞ=Qt ð3Þ

In the present experiments the relative heat loss e is found to be less than 5% for all cases.

The average single-phase liquid convection heat transfer coeﬃcient over the chip is deﬁned as

hl¼

Q_e

Achip ðTchip TinÞ

ð4Þ

where Tinis the coolant temperature at the inlet of the test

section and Tchip is the average temperature of the upper

surface of the chip which is estimated from the measured data for the lower surface of the chip by accounting for the 1-D heat conduction across the chip.

On the other hand, the average two-phase heat transfer coefficient for the coolant flow over the silicon chip is defined as

hr¼

Q_e

Achip ðTchip TsatÞ

ð5Þ

where Tsatis the saturated temperature of the coolant

FC-72.

Uncertainties of the single-phase liquid convection and ﬂow boiling heat transfer coeﬃcients and other parameters are estimated by the procedures proposed by Kline and

(7)

McClintock[23]. The detailed results from this uncertainty

analysis are summarized inTable 1.

4. Results and discussion

The present experiments are carried out for the FC-72

mass ﬂux G varying from 287 kg/m2s to 431 kg/m2s and

the imposed heat ﬂux q00 _{from 0.1 W/cm}2

to 10 W/cm2.

Besides, three silicon chips with the smooth, pin-ﬁnned 200 and pin-ﬁnned 100 surfaces are tested. The coolant is

at a pressure slightly lower than 1 atm with Tsat= 54.3°C.

Selected data are presented here to illustrate the effects of the coolant mass flux, imposed heat flux, and surface micro-structures on the saturated flow boiling heat transfer performance and associated bubble characteristics. The heat transfer performance is presented in terms of the boil-ing curves and boilboil-ing heat transfer coefficients.

4.1. Single-phase liquid convective heat transfer

Before beginning the boiling experiments, single-phase convective heat transfer tests are conducted for liquid FC-72. The measured mean single-phase liquid convection heat transfer coeﬃcients for the chip with a smooth surface are compared with the correlation proposed by Gersey and

Mudawar[1]for FC-72. Their correlation is

Nul¼ 0:362 Re0:614l Pr 1=3 _ð6Þ where Rel¼GL_l l and hl¼ kl L Nul.

The comparison is shown inFig. 4a for the dimensional

heat transfer coeﬃcients. The results indicate that our data are in a reasonable agreement with their correlation. Then the present data for the chips with the pin-ﬁnned surfaces

are given in Fig. 4b. Note that a signiﬁcant enhancement

in the liquid convection heat transfer can be obtained by adding the micro-pin-ﬁns to the chip surface. The enhance-ment is more pronounced at a higher ﬂow rate (Reynolds number). Besides, the chip surface with the smaller and

denser ﬁns shows a better heat transfer performance. The present data for the liquid convection from the pin-ﬁnned surfaces can be correlated as

Nul¼ 0:33 Re0:64l Pr

1=3_F

sp ð7Þ

with the factor Fsp accounting for the geometry eﬀects of

the ﬁns and it can be correlated as

Fsp¼ Sf H 0:15 _H Bf Wf 0:06 NAf Achip 0:04 ð8Þ Table 1

Summary of the uncertainty analysis

Parameter Uncertainty

Rectangular channel geometry

Length, width and thickness (%) ±0.5%

Area (%) ±1.0%

Parameter measurement

Temperature, T (°C) ±0.2

Temperature diﬀerence, DT (°C) ±0.3

System pressure, P (kPa) ±2

Mass ﬂux of coolant, G (%) ±2

Single-phase heat transfer in rectangular channel

Imposed heat ﬂux, q00(%) ±4.2

Heat transfer coeﬃcient, hl(%) ±12.3

Saturated ﬂow boiling heat transfer in rectangular channel

Imposed heat ﬂux, q00_(%) _±4.2

Heat transfer coeﬃcient, hr(%) ±12.3

200 300 400 500 600 G (kg/m2_s) 0 500 1000 1500 2000 hl (W/m 2K)

(a) FC-72 Single-phase Liquid Convective Heat Transfer at Tin = 25 ºC

- Smooth Surface Chip

- Correlation from Gersey & Mudawar (1992)

200 300 400 500 600 G (kg/m2_s) 0 500 1000 1500 2000 2500 hl (W/m 2K)

(b) FC-72 Single-phase Convective Heat Transfer at Tin = 25 ºC - Pin-finned100 Surface

- Pin-finned200 Surface - Smooth Surface — - Proposed Correlation —

Fig. 4. Comparison of the present single-phase liquid convection heat transfer coeﬃcient for the chip with (a) a smooth surface with the correlation of Gersey and Mudawar[1] and (b) micro-pin-ﬁns with the proposed correlation.

(8)

Here Sfis the pitch of the ﬁn array, H the rectangular

chan-nel height, Bfthe ﬁn height, Wfthe ﬁn width, N the total

number of pin-ﬁns, and Af is the surface area of a single

ﬁn. It should be mentioned that all the working ﬂuid

prop-erties used in reducing the data forFig. 4and Eqs.(6)–(8)

are evaluated at Tin.

4.2. Saturated ﬂow boiling curves

The eﬀects of the coolant mass ﬂux and surface micro-structure of the heated silicon chip on the boiling curves

are shown in Fig. 5. Note that for given G and surface

structure the chip surface temperature increases gradually

with the imposed heat ﬂux at a low q00 _{from the saturated}

temperature of the coolant to a certain value just exceeding

Tsatand no bubble nucleation is observed. The heat

trans-fer in this region is completely due to the single-phase

forced convection. With the continuing increase in q00_,

bub-bles begin to appear on the surface and the boiling curve is characterized by a sharp increase in the surface heat flux for a small rise in the temperature of the chip surface. We have onset of nucleate boiling (ONB) in the flow. The reason causing the transition in the boiling curve is due to a significant increase in the surface heat transfer by the boiling processes. It is also noted by comparing

the data in Figs. 5a and b that beyond ONB the coolant

mass ﬂux has rather slight eﬀects on the boiling curves,

sug-0 10 15 20 Δ Tsat (K) 0 2 4 6 8 10 12 14 16 q" (W /cm 2)

(a) FC-72 Saturated Flow Boiling Curves for G = 287 kg/m2_s - Pin-finned 100 Surface

- Pin-finned 200 Surface - Smooth Surface

ONB: Onset of Nucleate Boiling

0 10 15 20 0 2 4 6 8 10 12 14 16 q" (W /cm 2)

(b) FC-72 Saturated Flow Boiling Curves for G = 431 kg/m2_s - Pin-finned 100 Surface

ONB: Onset of Nucleate Boiling ONB ONB ONB ONB ONB ONB Δ Tsat (K) 5 5

Fig. 5. Boiling curves for various micro-structures of chip surface at (a) G = 287 kg/m2_{s and (b) G = 431 kg/m}2_s. 0 10 12 q" (W/cm2₎ 0 2000 4000 6000 8000 10000 h r (W/m 2K)

(a) FC-72 Saturated Flow Boiling Heat Transfer Coefficients for G = 287 kg/m2_s - Pin-finned 100 Surface - Pin-finned 200 Surface - Smooth Surface q" (W/cm2₎ 0 2000 4000 6000 8000 10000 h r (W/m 2K)

(b) FC-72 Saturated Flow Boiling Heat Transfer Coefficients for G = 431 kg/m2_s - Pin-finned 100 Surface - Pin-finned 200 Surface - Smooth Surface 2 4 6 8 0 2 4 6 8 10 12

Fig. 6. Saturated ﬂow boiling heat transfer coeﬃcients for various micro-structures of chip surface at (a) G = 287 kg/m2s and (b) G = 431 kg/m2s.

(9)

gesting that the surface heat transfer is mainly dominated by the fully developed nucleate boiling. Besides, at a higher G the required heat ﬂux for ONB is higher and this implies that more energy is needed for the vapor to nucleate from the wall since the residence time of the coolant on the chip

is shorter. The results in Fig. 5 further indicate that for

given G in both the single-phase and nucleate boiling regions at the same wall superheat the chip surface heat flux is highest for the pin-finned 100 surface and lowest for the smooth surface, showing that using the micro-pin-finned structure can effectively enhance the single-phase and flow boiling heat transfer from the chip surface. The large increase in the total surface area by adding the micro-pin-fins is indeed beneficial for the single- and two-phase heat transfer. Moreover, the wall superheat required for the boiling inception is substantially lower for the pin-finned 100 surface. This is attributed to the increase in the density of the active nucleation sites by the surface micro-structures. Particularly, in the corner region of the pin-fins

the bubbles are found to appear at lower DTsat.

4.3. Saturated ﬂow boiling heat transfer coeﬃcient

We continue to explore how the saturated ﬂow boiling

heat transfer coeﬃcient hr is aﬀected by the FC-72 mass

ﬂux and surface micro-structure on the silicon chips. The

data presented in Fig. 6 for the variations of hr with the

surface heat ﬂux reveal that for the three tested surfaces

the coolant mass ﬂux shows negligible inﬂuences on hr.

However, for a given coolant mass flux G the boiling heat transfer coefficient increases substantially with the imposed heat flux. Moreover, adding the surface micro-structures in the form of micro-pin-fins to the silicon chip produces some positive effects on the boiling heat transfer coefficient at high imposed heat flux. This is due to the fact that the boiling heat transfer performance can be raised effectively by the addition of active nucleation sites from the pin-fin structures. Besides, the heat transfer is further enhanced by a serpentine motion of liquid between adjacent rows of pin-fins, which in turn enhances turbulent mixing

[16,17]. Furthermore, development of multiple thermal entry regions at the top surfaces of individual pin-ﬁns is beneﬁcial.

4.4. Bubble characteristics

To elucidate the saturated flow boiling heat transfer characteristics presented above, the bubble characteristics obtained from the present flow visualization are examined in the following. At first, the top views of the boiling flow for various coolant mass fluxes and imposed heat fluxes are

shown inFig. 7. Bubbles begin to appear as the chip

sur-face temperature exceeds the incipient boiling superheat. In the beginning, tiny bubbles are seen on the active

(10)

ation sites. The bubbles grow and then detach from the

chip surface. As q00 _{increases, more bubbles are generated}

at more active nucleation sites and more bubbles detach from the chip surface. Besides, the detached bubbles tend to merge into larger bubbles, which occurs more frequently

at a higher q00_{. Note that the large bubbles become}

dis-torted and elongated when moving downstream. To quan-tify the bubble characteristics, the data for the mean bubble departure diameter and frequency and active nucleation

site density are given inFigs. 8–13.

The results in Fig. 8indicate that the mean size of the

bubbles departing from the chips can be reduced substan-tially by increasing the coolant mass flux. It is ascribed to the fact that the coolant at a higher mass flux and hence at a higher speed tends to sweep the bubbles more quickly away from the heating surface. Besides, the addition of the micro-pin-fins to the chips is found to cause a significantly

earlier departure of the bubbles and results in a smaller dp

(Fig. 9). The partially accelerated coolant flow between the fins can increase the drag force on the bubbles and sweep them away from the finned surface earlier. Note that for a given G the mean bubble departure diameter for the pin-finned 100 surface is larger than that for the pin-finned 200 surface. This due to the fact that the spaces between the

0 2 4 6 8 10 12 q" (W/cm2₎ 0 100 200 300 400 dp (μ m) dp (μ m) dp (μ m)

(a) FC-72 Saturated Boiling - Bubble Departure Diameters for smooth surface

- G = 287 kg/m2_s - G = 431 kg/m2_s 0 2 4 6 8 10 12 0 100 200 300 400

(b) FC-72 Saturated Boiling - Bubble Departure Diameters for pin-finned 200 surface

- G = 287 kg/m2_s - G = 431 kg/m2_s 0 2 4 6 8 10 12 0 100 200 300 400

(c) FC-72 Saturated Boiling - Bubble Departure Diameters for pin-finned 100 surface

- G = 287 kg/m2_s - G = 431 kg/m2_s

q" (W/cm2₎

Fig. 8. Mean bubble departure diameters for various coolant mass fluxes for the chips with (a) smooth surface, (b) pin-finned 200 surface and (c) pin-finned 100 surface. 0 2 4 6 8 10 12 q" (W/cm2₎ 0 100 200 300 400 dp (μ m)

(a) FC-72 Saturated Boiling - Bubble Departure Diameters for G = 287 kg/m2_s - Smooth Surface - G = Pin-finned 100 Surface - G = Pin-finned 200 Surface 0 2 4 6 8 10 12 q" (W/cm2₎ 0 100 200 300 400 dp (μ m)

(b) FC-72 Saturated Boiling - Bubble Departure Diameters for G = 431 kg/m2_s

- Smooth Surface

- G = Pin-finned 100 Surface - G = Pin-finned 200 Surface

Fig. 9. Mean bubble departure diameters for various micro-structures of chip surface at (a) G = 287 kg/m2s and (b) G = 431 kg/m2s.

(11)

adjacent fins for the pin-finned 100 surface are so small and the bubbles already contact the sides of the fins before departure. Therefore, the bubbles grow for a longer period of time before departure. Longer bubble growth time causes the growing bubble to keep absorbing energy from the heated surface for a longer period of time and results in a larger dp.

Next, the data given inFig. 10manifest that raising the

coolant mass ﬂux can signiﬁcantly augment the mean bub-ble departure frequency f. The increase of f with G is

ascribed again to the higher drag on the bubbles still attaching to the chip surface by the liquid coolant moving at a higher speed for a higher G. This, in turn, causes an earlier departure of the bubbles from the surface, resulting in a higher f. It is also observed that the mean bubble

departure frequency increases noticeably with q00_{. Finally,}

the data given in Fig. 11 indicate that f can be enhanced

effectively by using the pin-finned surface and the effect is slightly larger on the finned 200 surface than the pin-finned 100 surface.

Attention is turned to the data for the mean active

nucleation site density Nacshown in Figs. 12 and 13. Nac

is calculated based on the surface area of a bare chip.

Fig. 12manifests that an increase in the coolant mass ﬂux

0 2 4 6 8 10 12 q" (W/cm2₎ 0 500 1000 1500 2000 f (1/s)

(a) FC-72 Saturated Boiling - Bubble Departure Frequencies for smooth surface

- G = 431 kg/m2_s - G = 287 kg/m2_s 0 2 4 6 8 10 12 0 500 1000 1500 2000

(b) FC-72 Saturated Boiling - Bubble Departure Frequencies for pin-finned 200 surface

- G = 431 kg/m2_s - G = 287 kg/m2_s 0 2 4 6 8 10 12 0 500 1000 1500 2000

(c) FC-72 Saturated Boiling - Bubble Departure Frequencies for pin-finned 100 surface

- G = 431 kg/m2_s - G = 287 kg/m2_s f (1/s) f (1/s) q" (W/cm2₎ q" (W/cm2₎

Fig. 10. Mean bubble departure frequencies for various coolant mass fluxes for the chips with (a) smooth surface, (b) pin-finned 200 surface and (c) pin-finned 100 surface. 0 2 4 6 8 10 12 0 500 1000 1500 2000 f (1/s)

(a) FC-72 Saturated Boiling - Bubble Departure Frequencies for G = 287 kg/m2_s - Pin-finned 200 Surface - Pin-finned 100 Surface - Smooth Surface 0 2 4 6 8 10 12 0 500 1000 1500 2000 f (1/s)

(b) FC-72 Saturated Boiling - Bubble Departure Frequencies at for G = 431 kg/m2_s - Pin-finned 200 Surface - Pin-finned 100 Surface - Smooth Surface q" (W/cm2₎ q" (W/cm2₎

Fig. 11. Mean bubble departure frequencies for various micro-structures of chip surface at (a) G = 287 kg/m2_{s and (b) G = 431 kg/m}2_s.

(12)

results in somewhat lower Nacfor the three surfaces. This directly relates to the fact that the higher imposed heat ﬂux is needed for the boiling inception at a higher G, as already discussed in the previous section. Finally, it is noted from

Fig. 13 that adding the micro-pin-fins to the chip surface can effectively increase the active nucleation sites. This is due to the increase in the surface area of micro-pin-fins.

Moreover, Nac on the pin-ﬁnned 100 surface is slightly

higher than the pin-ﬁnned 200 surface.

4.5. Correlation equations

Based on the present experimental data, empirical corre-lations for the bubble characteristics and heat transfer coef-ficient in the saturated flow boiling of FC-72 on the heated silicon chip flush-mounted on the bottom of the rectangu-lar channel are proposed.

First, the data for the average bubble departure diame-ter can be correlated as

dp dp ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi r=g Dq p ¼0:25 ðql=qvÞ 0:48 Bo0:21

Re0:08_l for the smooth surface ð9Þ

0 2 4 6 8 10 12 0 10 20 30 40 50 Nac (10 5 n/m 2)

(a) FC-72 Saturated Boiling - Active Nucleation Site Densities for smooth surface

- G = 287 kg/m2_s - G = 431 kg/m2_s 0 2 4 6 8 10 12 0 10 20 30 40 50

(b) FC-72 Saturated Boiling - Active Nucleation Site Densities for pin-finned 200 surface

- G = 287 kg/m2_s - G = 431 kg/m2_s 0 2 4 6 8 10 12 0 10 20 30 40 50

(c) FC-72 Saturated Boiling - Active Nucleation Site Densities for pin-finned 100 Surface

- G = 287 kg/m2_s - G = 431 kg/m2_s q" (W/cm2₎ q" (W/cm2₎ q" (W/cm2₎ Nac (10 5 n/m 2) Nac (10 5 n/m 2)

Fig. 12. Mean active nucleation site densities for various coolant mass fluxes for the chips with (a) smooth surface, (b) pin-finned 200 surface and (c) pin-finned 100 surface. 0 2 4 6 8 10 12 0 10 20 30 40 50

(a) FC-72 Saturated Boiling - Active Nucleation Site Densities for G = 287 kg/m2_s - Pin-finned 100 Surface - Pin-finned 200 Surface - Smooth Surface 0 2 4 6 8 10 12 0 10 20 30 40 50

(b) FC-72 Saturated Boiling - Active Nucleation Site Densities for G = 431 kg/m2_s - Pin-finned100 Surface - Pin-finned200 Surface - Smooth Surface Nac (10 5 n/m 2) Nac (10 5 n/m 2) q" (W/cm2₎ q" (W/cm2₎

Fig. 13. Mean active nucleation site densities for various micro-structures of chip surface at (a) G = 287 kg/m2_{s and (b) G = 431 kg/m}2_s.

(13)

0 0 .1 0.2 0 .3 0.4 0 .5 d (Proposed Correlation)p 0 0.1 0.2 0.3 0.4 0.5 d(Experimental Data)p

(a) FC-72 Saturated Flow Boiling from Silicon Chip - Pin-finned 100 Surface - Pin-finned 200 Surface - Smooth Surface + 20% - 20% 0 300 0 6 00 0 9 000 12000 f (Proposed Correlation) 0 3000 6000 9000 12000 f (Experimental Data)

(b) FC-72 Saturated Flow Boiling from Silicon Chip - Pin-finned 100 Surface - Pin-finned 200 Surface - Smooth Surface + 25% - 25% 0 0 .05 0 .1 0.15 0.2 0 .2 5 Nacdp2 (Proposed Correlation) 0 0.05 0.1 0.15 0.2 0.25 Nac dp 2 (Experimental Data)

(c) FC-72 Saturated Flow Boiling from Silicon Chip - Pin-finned 100 Surface

+ 30%

- 30%

Fig. 14. Comparison of the measured data with the proposed correlations in the saturated ﬂow boiling of FC-72 for (a) mean bubble departure diameter, (b) mean bubble departure frequency and (c) mean active nucleation site density.

Y.M. Lie et al. /Internation al Journal of Heat and Mass Transfer 50 (2007) 3862–3876

(14)

and dp dp ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi r=g Dq p ¼0:15 ðql=qvÞ 0:46 Bo0:24_F d;sat Re0:07_l

for the pin-finned surfaces ð10Þ

with the ﬁn-geometry factor Fd,satcorrelated as

Fd;sat¼ Sf H 0:12 H Bf Wf 0:05 N Af Achip 0:42 ð11Þ

Fig. 14a shows that the present data for dp fall within ±20% of the correlations given above. In addition, the data for the mean bubble departure frequency can be correlated as f f dp l_l=q_l Dh ¼ 0:65Re1:3 l Pr

0:7_Bo0:66 _{for the smooth surface}

ð12Þ and f f dp ll=ql Dh ¼ 0:21Re1:3 l Pr 0:72_Bo0:66_F f;sat

with the ﬁn-geometry factor Ff,satcorrelated as

Ff;sat¼ Sf H 0:12 H Bf Wf 0:4 N Af As 0:02 ð14Þ

Fig. 14b reveals that most of the present data for f dpfall within ±25% the above two equations. Moreover, correla-tions for the mean active nucleation site density are pro-posed as

Nac d2p¼ 75 Bo

0:84_Re0:15

l for the smooth surface

ð15Þ and

Nac d2p¼ 72 Bo

0:87_Re0:15

l Fn;sat for the pin-finned surfaces

ð16Þ

with the factor Fn,satcorrelated as

Fn;sat¼ Sf H 0:15 H Bf Wf 0:06 N Af As 1:15 ð17Þ

Note that Fd,sat, Ff,satand Fn,satare the dimensionless

fac-tors to include the ﬁn-geometry eﬀects on the bubble

char-acteristics. The comparison given in Fig. 14c shows that

more than 85% of the present experimental data for Nacfall

within ±30% of the above correlations.

Finally, the heat ﬂux to the boiling ﬂow q00_{is considered}

to be composed of two parts: one resulting from the bubble

nucleation q00

b and another due to the single-phase forced

convection q00 c. Thus q00¼ q00 bþ q 00 c ð18Þ Here q00

b can be estimated from the above correlations for

the bubble characteristics as

q00_b¼ qv Vb f Nac ilv ð19Þ

where qvis the vapor density, Vbis the mean volume of the

departing bubble deﬁned as4p

3 dp

2

3

, and ilvis the enthalpy

of vaporization. Besides, q00

c can be estimated from the

single-phase liquid convection as

q00_c ¼ E hl DTsat ð20Þ

where E is an enhancement factor to account for the agitat-ing motion of the bubbles which can enhance the sagitat-ingle- single-phase heat transfer. From the present data, E can be empir-ically correlated as E¼ 4:5 N0:5 conf Fr 0:15 l ð1 þ 280 BoÞ 1:8

for the smooth surface ð21Þ

and E¼ 2:8 N0:5 conf Fr 0:13 l ð1 þ 270 BoÞ 1:8 FE;sat

with the ﬁn-geometry factor FE,satcorrelated as

FE;sat¼ Sf H 0:16 _H Bf Wf 0:1 _N Af As 0:02 ð23Þ

The comparison shown in Fig. 15 indicate that nearly all

the present heat transfer data fall within ±25% of the above correlations with an average deviation of 8.9%. 5. Concluding remarks

An experiment has been carried out to investigate satu-rated flow boiling of FC-72 on a heated micro-pin-finned silicon chip flush-mounted on the bottom of a rectangular channel. The effects of the imposed heat flux, coolant mass

0 3 0 6 0 9 0 1 20 q" (kW/m2_{, Proposed Correlation)} 0 30 60 90 120 q" (kW/m 2, Experimental Data)

FC-72 Saturated Flow Boiling from Silicon Chip - Pin-finned 100 Surface

+ 25%

- 25%

Fig. 15. Comparison of the measured heat transfer data with the proposed correlation in the saturated ﬂow boiling of FC-72.

(15)

ﬂux and surface structure on the measured data have been examined in detail. Furthermore, empirical equations to correlate the measured data are proposed. The major results obtained here can be summarized as follows:

(1) The boiling incipience heat ﬂux is higher at a higher coolant mass ﬂux.

(2) The coolant mass ﬂux shows little inﬂuence on hr.

Besides, increasing the imposed heat flux significantly promotes boiling heat transfer. Moreover, addition of the micro-pin-fins to the chips is also beneficial. (3) The mean bubble departure diameter is reduced at

increasing mass flux but the departing bubbles are significantly larger at a higher imposed heat flux. The bubbles departing from the heated surfaces are found to be somewhat smaller for the pin-finned surfaces.

(4) The bubble departure frequency increases with the coolant mass flux and imposed heat flux. Besides, higher bubble departure frequency is found on the pin-finned surfaces.

(5) The active nucleation site density decreases at increasing coolant mass flux. However, an opposite trend results for an increase in the imposed heat flux and for the addition of the micro-pin-fin structures to the chip surfaces.

Acknowledgement

The ﬁnancial support of this study by the engineering division of National Science Council of Taiwan, ROC through the contract NSC 93-2212-E-009-005 is greatly appreciated.

References

[1] C.O. Gersey, I. Mudawar, Effects of orientation on critical heat flux from chip arrays during flow boiling, Trans. ASME J. Electr. Pack. 114 (1992) 290–299.

[2] T.C. Willingham, I. Mudawar, Forced-convection boiling and critical heat ﬂux from a linear array of discrete heat sources, Int. J. Heat Mass Transfer 35 (1992) 2879–2890.

[3] T.J. Heindel, S. Ramadhyani, F.P. Incropera, Liquid immersion cooling of a longitudinal array of discrete heat sources in protruding substrates: II – forced convection boiling, Trans. ASME J. Electr. Pack. 114 (1992) 63–70.

[4] W.R. McGills, V.P. Carey, B.D. Strom, Geometry effects on critical heat flux for subcooled convective boiling from an array of heated elements, Trans. ASME J. Heat Transfer 113 (1991) 463–471. [5] C.P. Tso, K.W. Tou, G.P. Xu, Flow boiling critical heat flux of FC-72

from ﬂush-mounted and protruded simulated chips in a vertical rectangular channel, Int. J. Multiphase Flow 26 (2000) 351–365.

[6] K.R. Samant, T.W. Simon, Heat transfer from a small heated region to R-113 and FC-72, Trans. ASME J. Heat Transfer 111 (1989) 1053– 1059.

[7] Y. Ma, J.N. Chung, A study of bubble dynamics in reduced gravity forced-convection boiling, Int. J. Heat Mass Transfer 44 (2001) 399– 415.

[8] R. Situ, Y. Mi, M. Ishii, M. Mori, Photographic study of bubble behavior in forced convection subcooled boiling, Int. J. Heat Mass Transfer 47 (2004) 3659–3667.

[9] C.P. Yin, Y.Y. Yan, T.F. Lin, B.C. Yang, Subcooled ﬂow boiling heat transfer of R-134a and bubble characteristics in a horizontal annular duct, Int. J. Heat Mass Transfer 43 (2000) 1885–1896. [10] C. Bang, Won-Pil Baek, S.H. Chang, A digital photographic study on

nucleate boiling in subcooled ﬂow for water and refrigerant 134A ﬂuids, in: Proceedings of ICONE10 10th International Conference on Nuclear Engineering Arlington, VA, April 14–18, 2002.

[11] S.H. Chang, I.C. Bang, Won-Pil Baek, A photographic study on the near-wall bubble behavier in subcooled ﬂow boiling, Int. J. Therm. Sci. 41 (2002) 609–618.

[12] C. Bang, S.H. Chang, Won-Pil Baek, Visualization of the subcooled ﬂow boiling of R-134a in a vertical rectangular channel with an electrically heated wall, Int. J. Heat Mass Transfer 47 (2004) 4349– 4363.

[13] R. Maurus, V. Ilchenko, T. Sattelmayer, Study of the bubble characteristics and the local void fraction in subcooled ﬂow boiling using digital imaging and analyzing techniques, Exp. Therm. Fluid Sci. 26 (2002) 147–155.

[14] R. Maurus, V. Ilchenko, T. Sattelmayer, Automated high-speed video analysis of the bubble dynamics in subcooled ﬂow boiling, Int. J. Heat Fluid Flow 25 (2004) 149–158.

[15] G.E. Thorncroft, J.F. Klausner, R. Mei, An experimental investiga-tion of bubble growth and detachment in vertical upﬂow and downﬂow boiling, Int. J. Heat Mass Transfer 41 (1998) 3857–3871. [16] D.E. Maddox, I. Mudawar, Single- and two-phase convective heat

transfer from smooth and enhanced microelectronic heat sources in a rectangular channel, Trans. ASME J. Heat Transfer 111 (1989) 1045– 1052.

[17] I. Mudawar, D.E. Maddox, Enhancement of critical heat ﬂux from high power microelectronic heat sources in a ﬂow channel, Trans. ASME J. Electr. Pack. 112 (1990) 241–248.

[18] H. Honda, H. Takamastu, J.J. Wei, Enhanced boiling of FC-72 on silicon chips with micro-pin-ﬁns and submicron-scale roughness, Trans. ASME J. Heat Transfer 124 (2002) 383–389.

[19] H. Honda, J.J. Wei, Effects of fin geometry on boiling heat transfer from silicon chips with micro-pin-fins immersed in FC-72, Int. J. Heat Mass Transfer 46 (2003) 4059–4070.

[20] H. Honda, J.J. Wei, Enhanced boiling heat transfer from electronic components by use of surface microstructures, Exp. Therm. Fluid Sci. 28 (2004) 159–169.

[21] C. Ramaswamy, Y. Joshi, W. Nakayama, W.B. Johnson, Eﬀects of varying geometrical parameters on boiling from microfabricated enhanced structures, Trans. ASME J. Heat Transfer 125 (2003) 103– 109.

[22] K.N. Rainey, S.M. You, S. Lee, Eﬀect of pressure, subcooling, and dissolved gas on pool boiling heat transfer from microporous, square pin-ﬁnned surface in FC-72, Int. J. Heat Mass Transfer 46 (2003) 23– 25.

[23] S.J. Kline, F.A. McClintock, Describing uncertainties in single-sample experiments, Mech. Eng. 75 (1953) 3–12.