International Journal of Heat and Mass Transfer 80 (2015) 26–37

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Experimental investigation of spray cooling on micro-, nano-and hybrid-structured surfaces

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2014.08.0850017-9310/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +86 10 62772661; fax: +86 10 62775566.E-mail address: jiangpx@tsinghua.edu.cn (P.-X. Jiang).

Zhen Zhang a,b, Pei-Xue Jiang a,⇑, David M. Christopher a, Xin-Gang Liang c

a Key Laboratory for Thermal Science and Power Engineering of the Ministry of Education, Department of Thermal Engineering, Tsinghua University, Beijing 100084, PR Chinab Institute of Nuclear and New Energy Technology, Tsinghua University, Beijing 100084, PR Chinac Key Laboratory for Thermal Science and Power Engineering of the Ministry of Education, School of Aerospace Engineering, Tsinghua University, Beijing 100084, PR China

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

Article history:Received 28 May 2014Received in revised form 27 August 2014Accepted 30 August 2014

Keywords:Spray coolingMicro-Nano-Hybrid-structured surfaceLocal thermal non-equilibrium model

The heat transfer during spray cooling was studied experimentally using deionized water to investigatethe spray characteristics and the differences between spray cooling on a smooth silicon surface andmicro-, nano- and hybrid micro/nano-structured surfaces. The spray cooling experiments show thatthe heat transfer rates were better for the nano-structured surface, followed by the smooth surfacecoated with the SiO2 film and the pure silicon surface since the contact angle was smallest on thenano-structured surface and increased on the latter two. The droplet parameter results show that mostdroplets were 40–60 lm in size. The heat transfer coefficient increased and the wall temperaturedecreased on the 25G � 25S surface coated with the SiO2 film compared with the 50G � 50S surfacecoated with the SiO2 film as the heat transfer moved into the partial dryout region due to the SiO2 film’sstronger hydrophilicity so the heated area was more fully utilized, while the CHF was larger for the50G � 50S surface. Coating the micro-structured surfaces with carbon nano-tube (CNT) films havingcharacteristic sizes smaller than the droplet size was more effective than on the surfaces with larger char-acteristic sizes. The CHF was largest on the 25G � 25S surface coated with 4 carbon nano-tube films witha 75.3% increase over the smooth surface. The wall temperature increase and the temperature fluctua-tions were small in the boiling regime as the power increases for the enhanced surfaces.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Numerous industrial applications, such as electronic systems,high-energy lasers, energy weapons and aerospace satellites, havesubstantial need for more effective thermal management. Spraycooling, with its high heat dissipating capability, has been playingan important role in high heat flux applications as one of the mosteffective thermal management methods. Heat fluxes in excess of1000 W/cm2 have been reported using water spray cooling atlow coolant flow rates [1].

Spray cooling heat transfer is influenced by many factors suchas the droplet parameters [2], working fluid [3], nozzle-to-surfacedistance and inclination angle [4,5], so it has been widely investi-gated by researchers in the past two decades. Surface morphologyis another critical effect for spray cooling heat transfer enhance-ment. Enhanced surfaces, such as milli-structured surfaces [6],micro-structured surfaces [7,8] and microcavity surfaces [9,10]have been shown to effectively improve the heat transfer.

Spray cooling heat transfer on smooth and micro-structuredsurfaces (characteristic sizes of 25–200 lm) has been studied indetail in previous research in this group [11] with the resultsshowing that micro-structured surfaces effectively increased thespray cooling heat transfer rates in the thin film and partial dryoutregions. The effects of the groove width and stud size on the heattransfer were correlated with the droplet parameters. The micro-structured surface with larger characteristic sizes had a smallerarea enhancement factor and worse heat transfer rates, while onthe micro-structured surface with much smaller characteristicsizes, most droplets were not able to completely enter the bottomof the micro grooves since the groove size was smaller than most ofthe droplets and the heat transfer surface could not be fully wetted.However, the spray cooling heat transfer on nano- or hybrid micro/nano-structured surfaces is not clearly understood and the litera-ture has few studies on their effects to the authors’ knowledge.

Only Alvarado [12] has investigated spray cooling on nano-structured surfaces and observed lower minimum wall tempera-tures for similar heat fluxes, better heat transfer curves, and lowertemperature gradients on the nano-structured surfaces than on thebare surface. Nano-structured surfaces have also been used for

Nomenclature

G groove width, lmS stud size, lmD groove depth, lmTW silicon top surface temperature, �CTPt platinum temperature, �CTf water temperature, �CT average temperature in an x cross section of the silicon,

�Cq heat flux, W/cm2

h heat transfer coefficient, W/(m2 �C)H silicon thickness, lmHc critical thickness, lmL length, lme volume fraction, ei = Li/(Lh + Luh)k thermal conductivity, W/(m �C)R100 standard resistance, XRx platinum resistance, XU100 standard resistance voltage, VUx platinum voltage, VQx electrical heat source in the platinum heater

in each cell, W

Qelec total electrical heat source on the silicon, WQ heat dissipated by the spray, WD10 average diameter, lmD32 Sauter mean diameter, lmv droplet velocity, m/sAS stud top surface area directly impinged by the water

spray, mm2

AB base area, 7.4 mm � 7.4 mmAT total area of the micro-structured surface, mm2

AS/AB directly impinged surface percentageAT/AB area enhancement factor compared to the smooth

surface

Superscriptsh heateduh unheatedc critical0 silicon bottom conditionssi silicons loss

Z. Zhang et al. / International Journal of Heat and Mass Transfer 80 (2015) 26–37 27

pool boiling with the results helping understanding of the spraycooling heat transfer on such surfaces. Young Lee et al. [13] inves-tigated the nucleate pool boiling heat transfer and long-term per-formance of a nano-porous surface fabricated by anodizing withthe results showing that the nucleate boiling heat transfer coeffi-cient of the nano-porous coating surface was higher than that ofthe non-coated surface particularly at low heat fluxes with higherheat transfer coefficients remaining throughout 500 h of operation.Im et al. [14] fabricated copper nanowire arrays on a silicon sub-strate by electro-chemical deposition and observed that the coppernanowires increased the pool boiling Critical Heat Flux (CHF) andreduced the wall superheat compared to a smooth silicon surface.An optimum CHF was found at a nanowire height of 2 lm. Kwarket al. [15] found that a nanoparticle coated heater consistently pro-duced dramatic CHF enhancement relative to an uncoated surfacefor all tested conditions. The authors postulated that the betterwettability in the nanocoating was the main cause of the enhance-ment. Thus, all the experiments show that the nano-structures onthe surface improve the heat transfer rates.

In addition, hybrid micro/nano-structured surfaces have beenused in pool boiling to further enhance the heat transfer in recentyears. Various micro-, nano- and hybrid-structures were fabricatedon copper surfaces with the corresponding pooling boiling heattransfer performance studied by Li et al. [16]. The authors claimedthat the CHF of the hybrid-structured surfaces was about 15%higher than that of the surfaces with nanowires only andmicro-pillars only, and that the superheat at CHF for the hybrid-structured surface was about 35% smaller than that of the micro-pillared surface. Launay et al. [17] studied pool boiling on smooth,nano-, micro- and hybrid-structured surfaces using PF5060 andwater. The highest heat fluxes were obtained using the 3D micro-structures without CNTs and the experimental results indicate thatthe heat transfer rates were higher on the purely nano-structuredinterfaces only at very low superheats compared to the smoothsurfaces, but still lower than those on the conventional Si-etchedmicrostructures for all cases.

The primary objective of the current study is to investigatespray cooling heat transfer on smooth and nano-, micro- andhybrid-structured surfaces with accurate measurements of thespray droplet parameters using the shadowgraph technique toexplain the heat transfer mechanism.

2. Experimental system and parameter measurements

2.1. Experimental system

The spray cooling system shown in Fig. 1 included spray, heat-ing and measurement sections. Deionized water driven by a Fluid-o-Tech magnetic drive gear pump flowed from the constant tem-perature water bath through the filter to remove impurities beforebeing sprayed on the heated surface through a full cone pressureatomizer (Spraying Systems Co.) with a nozzle orifice of0.51 mm. The nozzle was fixed in a bracket with the orifice-to-sur-face distance adjusted by an accurate micrometer with a position-ing accuracy of 0.01 mm. A mechanical pressure gauge was usedto measure the nozzle inlet pressure which was assumed to beequal to the spray pressure with one OMEGA 0.125 mm diameterT-type thermocouple imbedded in the flow tube just before thenozzle to measure the deionized water temperature. The experi-ments used a water spray pressure of 0.3 MPa with an orifice-to-surface distance of 30 mm with subcoolings of �82 to �80 �C.A flow rate measurement container with a square hole the samesize as the heated surface was made to measure the water flowrate impinging the target surface as in Ref. [11]. The flow ratewas 1.239 kg/m2 s in all cases.

The heating sections were made of 7.4 mm � 7.4 mm, double-side polished, 490 lm thick silicon dies. One CNT film or fourCNT films laid in the cross direction with tube diameters of 80–90 nm were laid on the top surface. Then, Plasma Enhanced Chem-ical Vapor Deposition (PECVD) was used to deposit a SiO2 film hav-ing a thickness 50 nm on the CNT films to prevent the nanotubesfrom being washed away by the spray. These surfaces are referredto as nano-structured surfaces. A smooth surface having just a50 nm thick SiO2 film, called the SiO2 film surface, was used tostudy the wettability effect in comparison with the smooth surfaceand to study the nano-scale structures effect compared with thenano-structured surfaces.

A three-dimensional white light surface interference profilome-ter was used to measure the roughnesses of the smooth surfaceswith a resolution of 752 � 489 pixels, RMS repeatability precisionof 1 nm and calibration accuracy� 0.1%. Sa was defined as theaverage difference in the profile height from the average heightwithin the sample length, which is widely used to characterize

Fig. 1. Spray cooling system.

(a)

(b) Smooth +1CNT (c) Smooth +4CNT

Fig. 2. Roughness distribution on the smooth surface (a) and surface appearances observed by an atomic force microscope on the nano-structured surfaces (b) and (c).

28 Z. Zhang et al. / International Journal of Heat and Mass Transfer 80 (2015) 26–37

the surface roughness. The roughness distribution for the smoothsurface is shown in Fig. 2(a) with Sa = 0.451 nm. The nano-struc-tured surface appearances were observed by an atomic forcemicroscope as shown in Fig. 2(b) and (c).

Various size micro-studs were fabricated on the top surface ofthe smooth silicon dies by deep reactive ion etching (DRIE). FourCNT films were then laid in the cross direction on the top surfacewith the CNT films between the micro-studs ablated by a laser toensure that the spray could flow into the grooves. Thus, the CNTfilms were only deposited on the top surface of the micro-studs.

A PECVD SiO2 film was also deposited on the micro-structured sur-faces with the micro-structured silicon dies just having a SiO2 filmfor reference. Ten test surfaces were investigated with their sizeparameters listed in Table 1.

Four platinum resistors were designed on the bottom surface ofeach silicon die to reduce the voltage input for safety consider-ations with each platinum resistance being 300–400 X. Chromium,platinum and titanium (thickness proportions of approximately1:10:1) were applied to the bottom surface of the silicon dies usingpositive photoresist lift-off in a serpentine pattern with a total

Table 1Test surface parameters.

Surface Abbreviation in the analysis

1 Smooth silicon Smooth2 Smooth silicon + a SiO2 film Smooth + SiO2

3 Smooth silicon + one CNT film + a SiO2 film Smooth + 1CNT4 Smooth silicon + four criss-crossed CNT films + a SiO2 film Smooth + 4CNT5 25G � 25S � 100Da + a SiO2 film 25G � 25S + SiO2

6 25G � 25S � 100D + four criss-crossed CNT films + a SiO2 film 25G � 25S + 4CNT7 50G � 50S � 100D + a SiO2 film 50G � 50S + SiO2

8 50G � 50S � 100D + four criss-crossed CNT films + a SiO2 film 50G � 50S + 4CNT9 100G � 100S � 100D + a SiO2 film 100G � 100S + SiO2

10 100G � 100S � 100D + four criss-crossed CNT films + a SiO2 film 100G � 100S + 4CNT

a G = groove width, lm, S = stud size, lm, D = groove depth, lm. 25G � 25S � 100D means that the groove width between the studs is 25 lm, the stud size is 25 lm and thegroove depth is 100 lm.

HLh Luh

q0

hT

0hT 0uhT

x

y

Unit cell

0

wT

Hc

qh0

uhT

quh0

Fig. 3. Silicon temperature profile in a unit cell.

Z. Zhang et al. / International Journal of Heat and Mass Transfer 80 (2015) 26–37 29

thickness of 241 nm. A PECVD SiO2 film having a thickness 150 nmwas also added on the metals for protecting. The SiO2 film and tita-nium on just the eight pads were removed by wet etching toexpose the platinum before the silicon die was mounted on a tem-perature resistant PCB circuit board. Eight 40 lm gold wires con-nected the four platinum resistors with the circuit board by wirebonding. A DC stabilized voltage source was then used to supplypower to the platinum through eight wires soldered on the circuitboard.

A synthetic glass sleeve filled with calcium silicate cellucotton(thermal conductivity = 0.05 W/mK) was glued to the bottom ofthe circuit board to reduce the heat losses. The system was consid-ered to be at steady state after the voltage and the temperatureremained constant for at least 15 min with the results calculatedusing the last 5 min.

2.2. Temperature and heat flux measurements

The platinum was used both as the heat source and a resistancethermometer to measure the bottom surface temperature that wasthen extrapolated to the top silicon surface temperature due to itslinear resistance–temperature relationship over a wide tempera-ture range with high repeatability and chemical stability. The heatflux uniformity in the silicon was guaranteed by the serpentineplatinum heaters on the bottom surface being very closely spaced[11]. Four 100 X BZ3 high power standard resistances with preci-sions of 0.01 X were connected in series with the four platinumheaters. The platinum resistance was then calculated from thevoltage ratio and the standard resistance as Rx = (Ux/U100)R100.The electrical heat into the platinum in each cell, Q x ¼ U2

x=Rx,and the total electrical heat into the silicon, Q elec =

Px=1�4Q x, were

used to calculate the heat dissipated by the spray, Q = Qelec � Qs.Thus, the heat flux on the silicon top surface was calculated as

q ¼ Qelec � Q s

L2si

¼P

x¼1�4U2x=Rx � Q s

L2si

ð1Þ

where Q s is the heat loss obtained from a polynomial curve of themeasured heat losses correlated with the temperature differencebetween the test surface and the surroundings without spray cool-ing and Lsi is the length of the silicon, 7.4 mm.

The relationship between the temperature and the platinumresistance was calibrated before the spray experiments in a HARTSCIENTIFIC 6022 water bath with thermostatic control for every5 �C between 25 and 85 �C with a precision of 0.01 �C, and the plat-inum temperature was then calculated from this calibration.

The top silicon surface temperature could be extrapolated fromthe platinum heater temperature using the local thermal non-equi-librium model in a solid zone as shown in Fig. 3 which was firstproposed by Ouyang et al. [18] to study the heat transfer in the

impermeable wall bounded by a porous medium. Half of theheated and unheated sections were treated as a unit cell with sym-metry boundary conditions. The platinum thickness was approxi-mately 241 nm, quite small compared with the silicon thickness,H, 490 lm, thus, the platinum was modeled as a surface heatsource regardless of the temperature gradient across its thickness.Assume T(x, y) can be given by a third order profile,T(x, y) = A(x) + B(x)y + C(x)y2 + D(x)y3, which satisfies the followingconditions

ThðxÞ ¼R 0

�LhTdy

Lhy ¼ �Lh : @T

@y ¼ 0

TuhðxÞ ¼R 0

�LuhTdy

Luhy ¼ Luh : @T

@y ¼ 0

8>><>>: ð2Þ

The variables Th, Tuh, Lh, Luh, represent the average temperaturesin the x cross section and the lengths of the heated and unheatedsections. The parameters A, B, C and D were then determined fromthese four equations. Taking the smooth silicon as an example,Lh = 50 lm, Luh = 100 lm and the volume fractions eh = 1/3 andeuh = 2/3. On the face at y = 0,

q ¼ hðTh � TuhÞ ¼ �k@T@y

����y¼0

ð3Þ

Thus

h ¼ � kB

ðTh � TuhÞ¼ 12LhLuhk

ðLh þ LuhÞðL2h þ 3LhLuh þ L2

uhÞð4Þ

The volumetric heat transfer coefficient has the form

hv ¼h

ðLh þ LuhÞ¼ 12euhð1� euhÞkðLh þ LuhÞ2ð1þ euh � e2

uhÞð5Þ

For the heated and unheated sections

30 Z. Zhang et al. / International Journal of Heat and Mass Transfer 80 (2015) 26–37

ehkd2Th

dx2 � hvðTh � TuhÞ ¼ 0 ð6Þ

euhkd2Tuh

dx2 þ hvðTh � TuhÞ ¼ 0 ð7Þ

Assuming T 0 ¼ Th � Tuh and T ¼ ehTh þ euhTuh, Eqs. (6) and (7) can bewritten as

d2T 0

dx2 ¼hv

eheuhkT 0 ð8Þ

kd2T

dx2 ¼ 0 ð9Þ

Eqs. (8) and (9) can then be solved with the following boundaryconditions

x ¼ 0 : T 0 ¼ Th0 � Tuh0; T ¼ ehTh0 þ euhTuh0

x ¼ H : T ¼ TW

x!1 : T 0 � 0

8><>: ð10Þ

where Th0 and Tuh0 are the silicon bottom average temperatures ofthe heated and unheated sections. Then,

T 0 ¼ ðTh0 � Tuh0Þe�mx ð11Þ

T ¼ ðehTh0 þ euhTuh0Þ þTW � ðehTh0 þ euhTuh0Þ

H

!x ð12Þ

where

m ¼ 1ðLh þ LuhÞ

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi12

ð1þ euh � e2uhÞ

sð13Þ

As shown in Eq. (11), the temperature difference decreases expo-nentially with x. The non-equilibrium region thickness, the criticalthickness, Hc, can be defined as

x ¼ Hc : T 0 ¼ 0:01T 00 ð14Þ

The critical thickness is then

Hc ¼1m

ln 100 ð15Þ

For the experimental conditions, Hc = 220.5 lm according to Eq.(15), less than the silicon thickness, H of 490 lm; therefore, the sil-icon top surface temperature can be regarded as uniform.

An energy balance in the unit cell gives

q ¼ ehqh0 þ euhquh0 ¼ �k ehdTh

dx

�����x¼0

þ euhdTuh

dx

�����x¼0

!¼ �k

dTdx

�����x¼0

ð16Þ

qh0 � quh0 ¼ �kdTh

dx

�����x¼0

� dTuh

dx

�����x¼0

!¼ �k

dT 0

dx

�����x¼0

; quh0 ¼ 0 ð17Þ

The silicon top surface average temperature, TW , can be calcu-lated from Eqs. (16) and (17) as

TW ¼ Th0 �qHk� euhq

kð1� euhÞmð18Þ

The silicon top surface temperature was then calculated usingEq. (18). The silicon top surface temperature was also calculatedby ANSYS Fluent 13.0 for validation (half of the heated andunheated sections were treated as a unit cell with symmetry

boundary conditions) as in reference [19], and the results showthat the differences between top surface temperature calculatedby Eq. (18) and Fluent were neglected.

The heat transfer coefficient was calculated as h = q/(TW � Tf),where Tf is the spray inlet temperature.

2.3. Contact angle measurement

The contact angles of deionized water on the various surfaceswere measured before the spray cooling experiments using anEasydrop (Kurss, Germany) with a precision of 0.1� as shown inTable 2. The results showed that the silicon surface was hydro-philic with a water contact angle of 62.5�. The silicon surfacecoated with the SiO2 film had a contact angle of 13.1�, while thenano-structured surfaces had similar wetting characteristics withcontact angles of 10.5� and 8.0�.

2.4. Droplet parameter measurements

In spray cooling, the droplets produced by the spray nozzle haveunique parameters, such as the droplet size distribution, dropletnumber density, and velocities that change with the liquid flowrate and nozzle type. The characteristic sizes of the micro-struc-tured surfaces in the present experiments were 25–100 lm, simi-lar to the spray droplet diameters, with both affecting the heattransfer rates. Thus, accurate measurements of the droplet charac-teristics were needed.

The shadowgraph technique using backlighting by a PIV systemmanufactured by LAVISION was used to measure the spray dropletparameters. The measurement procedure and principle, dropletsize calculation method and system verification were describedpreviously [11]. The results show that D10 = 55.4 lm,D32 = 103.4 lm, r = 32.0 lm and v = 9.65 m/s with most dropletsbeing 40–60 lm in size.

2.5. Experimental uncertainty analysis

The flow rate was obtained from an electronic scale with anuncertainty of 0.01 g and a stopwatch with an uncertainty of0.1 s, so the uncertainty of the flow rate was ±0.02%. The uncer-tainty in the voltage drop across the platinum, Ux, was 0.0073%and the voltage uncertainty of the standard resistance, U100, was0.016% for the voltages measured by the 34970A data acquisitionsystem. Thus, the platinum resistance uncertainty was

dRx

Rx

�������� ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffidUx

Ux

� �2

þ dU100

U100

� �2

þ dR100

R100

� �2s

¼ 0:020% ð19Þ

The uncertainty in the electrical heat into the platinum in each cellwas then

dQx

Q x

�������� ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffidUx

Ux

� �2

� 2þ dRx

Rx

� �2s

¼ 0:023% ð20Þ

The uncertainty of the total electrical heat into the silicon from thefour platinum electrical heaters was then 0.012%.

A heat loss of less than 2.86% of the total input power wasassumed to be the heat loss uncertainty. The silicon length uncer-tainty, dLsi, was 40 lm since the scribing mark widths were 20 lmon each side for the manufacturing process. Thus, from Eq. (1), theheat flux uncertainty was calculated to be:

dqq

�������� 6

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffidQ elec

Q elec

� �2

þ dQ s

Q elec

� �2

þ 2� dLsi

Lsi

� �2s

¼ 3:06% ð21Þ

Table 2Deionized water contact angles on various surfaces.

Surface Droplet shape Contact angle(�)

Si (smooth) 62.5

Si + SiO2 13.1

Si + 1CNT + SiO2 10.5

Si + 4CNT + SiO2 8.0

Z. Zhang et al. / International Journal of Heat and Mass Transfer 80 (2015) 26–37 31

The relationship between the temperature and the platinumresistance was calibrated using a water bath with a precision of0.01 �C, so the temperature uncertainty in the second-order form,dT1, was ±0.08 �C. The maximum temperature uncertainty causedby the platinum resistance uncertainty according to the second-order fit, dT2, was ±0.17 �C. Thus, the platinum temperature uncer-tainty, dTPt, was ±0.19 �C.

The silicon top surface temperature was calculated usingEq. (18). The uncertainty in the platinum width, spacing and siliconthickness were negligible due to the precise micro-manufacturingprocess. Thus, the uncertainty of the top surface temperature wascalculated to be

jdTwj ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðdTPtÞ2 þ

Hkþ euh

kð1� euhÞm

� �2

ðdqÞ2s

¼ 0:32 �C ð22Þ

The uncertainty of the heat transfer coefficient was:

dhh

�������� ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffidqq

� �2

þ dTw

Tw � Tf

� �2

þ dTf

Tw � Tf

� �2s

¼ 3:6% ð23Þ

3. Results and discussion

3.1. Spray cooling on smooth and nano-structured surfaces

Fig. 4 shows the heat fluxes while Table 3 shows the heated sur-faces at various heat fluxes on the smooth silicon surface, thesmooth surface coated with the SiO2 film and the nano-structuredsurfaces. The SiO2 film surface, which had a contact angle of just13.1�, was more hydrophilic than the smooth silicon surface witha contact angle of 62.5�. The smaller contact angle resulted in fasterliquid spreading and better heat transfer with a higher CHF on the

-5 0 5 10 15 20 25 3040

60

80

100

120

140

q (W

/cm

2 )

Tw - T

sat (0C)

Smooth Smooth + SiO

2

Smooth + 1CNT Smooth + 4CNT

Fig. 4. Heat fluxes on the smooth and nano-structured surfaces.

SiO2 film surface than on the smooth surface as shown in Fig. 4. Thenano-structured surfaces had even better hydrophilicity than theSiO2 film surface with contact angles of 10.5� and 8.0�. The nano-structures increased the heat transfer area and the spaces betweenthe carbon nano-tubes retained the liquid on the surface longer.The superheat degree was approximately 12.4 �C on the smoothsurface, 8.7 �C on the SiO2 film surface, 9.0 �C and 5.9 �C on thenano-structured surfaces at a heat flux of 80 W/cm2. The CHF onthe SiO2 film surface was about 5.8% higher than that on thesmooth surface, and the CHFs on the two nano-structured surfaceswere about 12.6% and 11.6% higher than that on the smooth sur-face. Therefore, the heat transfer was enhanced by higher wettabil-ity, although it was not quite significant.

The heat transfer regime successively undergoes four regions,the single phase region, the thin film region, the partial dryoutregion and the dryout region as in reference [11]. The three heatedsurfaces were covered with a thick liquid film as shown in picturesA1–C1 in Table 3 at low heat fluxes since the heat transfer was inthe single phase region. The wall temperature was not strictly uni-form since the spray droplet impacts were discrete events, but thetemperature fluctuations were small and slightly increased withthe heat flux. The heated surfaces were still completely coveredwith the liquid film in pictures A2–C2, while the film was thinnerthan in A1–C1 when the heat transfer entered the thin film region.Heat was dissipated largely by phase change with strong evapora-tion and the heat transfer curve slope increased greatly. The non-uniform effects caused by the discrete droplet impacts becamesmaller and the wall temperature fluctuations decreased as thehigh, uniform heat flux was dissipated.

The liquid film then began to rupture as the heat flux increasedwith the heat transfer entering the partial dryout region, with theportion of the liquid film that had ruptured becoming larger as theheat flux increased as shown in A3–A4 and B3–B4. The unwettedarea was much smaller on the SiO2 film surface than on the smoothsurface for the same heat flux. The heat transfer on the nano-struc-tured surface remained in the thin film region for heat fluxes up toapproximately 111 W/cm2 as shown in C3 and began to ruptureslightly when the heat flux reached 122 W/cm2 as shown in C4due to the stronger hydrophilicity which kept the surface wettedat higher heat fluxes. The temperature fluctuations increased rap-idly due to the unsteady, repeated breakup and recovery of theliquid film for all three surfaces, but the average temperatureremains the same in a long period. The system was considered tobe at steady state after the voltage and the average temperatureremained constant for at least 15 min with the results calculatedusing the last 5 min. The wall temperature increased further inthe dryout region until the entire surface was dry and Critical HeatFlux, CHF, was achieved in A5–C5. The wall temperature fluctuatedmuch more due to the extremely unsteady liquid film in thisregion.

3.2. Spray cooling on smooth and micro-structured surfaces

Fig. 5 shows the heat fluxes on the smooth and micro-struc-tured surfaces with the results showing that the micro-structuredsurfaces have higher spray cooling heat transfer rates than thesmooth surface in the thin film and partial dryout regions. Thisheat transfer enhancement is attributed to the micro-structureson the heating surface, which not only increased the heat transferarea but also provided capillary forces along the various surfaces asdriving forces to spread the deposited liquid film to keep the sur-face wetted at higher heat fluxes. The liquid distribution was thenmore uniform and the evaporation was faster, causing significantincreases of the heat transfer. Moreover, the micro-structures onthe surface also provide additional surface nucleation sites.

Table 3Heated surfaces at various heat fluxes on smooth and nano-structured surfaces.

A. Smooth B. Smooth + SiO2 C. Smooth + 4CNT

1

q = 19.0 W/cm2 q = 21.8 W/cm2 q = 23.0 W/cm2

Tw = 56.5 �C Tw = 58.5 �C Tw = 27.6 �C2

q = 79.8 W/cm2 q = 78.6 W/cm2 q = 80.1 W/cm2

Tw = 117.6 �C Tw = 115.1 �C Tw = 114.4 �C3

q = 109.7 W/cm2 q = 111.8 W/cm2 q = 110.9 W/cm2

Tw = 124.0 �C Tw = 123.4 �C Tw = 122.2 �C4

q = 121.0 W/cm2 q = 121.3 W/cm2 q = 121.5 W/cm2

Tw = 126.8 �C Tw = 125.8 �C Tw = 124.2 �C

32 Z. Zhang et al. / International Journal of Heat and Mass Transfer 80 (2015) 26–37

Table 3 (continued)

A. Smooth B. Smooth + SiO2 C. Smooth + 4CNT

5

q = 123.7 W/cm2 q = 130.9 W/cm2 q = 138.1 W/cm2

Tw = 128.1 �C Tw = 127.9 �C Tw = 128.2 �C

-5 0 5 10 15 20 25 30

60

90

120

150

180

210

Hea

t flu

xes

(W/c

m2 )

Tw - T

sat (0C)

Smooth Smooth + SiO

2

25G 25S + SiO2

50G 50S + SiO2

100G 100S + SiO2

Fig. 5. Heat fluxes on the smooth and micro-structured surfaces.

Z. Zhang et al. / International Journal of Heat and Mass Transfer 80 (2015) 26–37 33

Therefore, the micro-structured surfaces effectively improve thespray cooling heat transfer rates.

In the partial dryout region, the liquid film begins to rupture asthe heat flux increases so that the liquid can no longer completelywet the entire heated surface. For the micro-structured surface, themany micro-studs that puncture the liquid film greatly increasethe contact line length which greatly increases the heat transferrate. For instance, on the 25G � 25S + SiO2 surface, if each studcould be surrounded by the liquid in the limiting cases, the contactline length could be approximately 70 times larger than that on thesmooth surface. Sodtke and Stephan [20] also believed that theimproved spray cooling performance on the micro-structured sur-face compared to the smooth surface was due to an increasedlength of the three phase contact line that formed on the structureswhich led to a very efficient thin film evaporation. Horacek et al.[21] proposed that the wall heat flux correlated very well withthe contact line length as well. Therefore, the wall temperatureincreased slightly and then decreased a larger amount despitethe increased heat flux. This observation agrees well with the find-ings of Sodtke and Stephan [20] and Pais et al. [22].

Zhang et al. [11] stated that most of the droplets in this systemhad sizes of 40–60 lm, so the droplets could go into the groovesand take advantage of the enhanced area on the 50G � 50S � 100Dsurface. However, the groove width on the 25G � 25S � 100D sur-face was too small for the droplets to get to the bottom of thegrooves so the enhanced area could not be completely utilized,leading to poorer heat transfer rates than for the50G � 50S � 100D surface although the area enhancement factorof the 25G � 25S � 100D surface was 5.0, much larger than that

of the 50G � 50S � 100D surface, 3.0. (The area enhancement fac-tor is the ratio of the micro-structured surface total area to thesmooth surface area.)

The PECVD SiO2 film on the micro-structured surfaces made thesurfaces more hydrophilic with faster liquid spreading whichenabled more liquid to get into the small grooves than on the sili-con micro-structured surface without the SiO2 film. The heat trans-fer rates for the micro-structured surfaces coated with the SiO2

film were similar in the thin film region. In the partial dryoutregion, many micro-studs that punctured the liquid film greatlyincreased the contact line length which greatly increased the heattransfer rate with the wall temperature increasing slightly andthen decreasing a larger amount despite the increased heat flux.The contact line length and the heat transfer coefficient increaseon the 25G � 25S + SiO2 film surface got much more with the walltemperature decreasing more intensely compared with the50G � 50S + SiO2 film surface due to the SiO2 film’s stronger hydro-philicity and the heated area being more fully utilized, with theCHF still larger for the 50G � 50S + SiO2 film surface. The heattransfer rates on the 100G � 100S + SiO2 film surface with an areaenhancement factor of 2.0 were lower than those of the other twosurfaces with a lower CHF.

3.3. Spray cooling on micro-structured and hybrid-structured surfaces

Fig. 6 compares the heat fluxes on the 25G � 25S micro- andhybrid-structured surfaces. The heat transfer performance wasbetter on the hybrid-structured surface than on the micro-struc-tured surface with higher heat fluxes as the heat transfer regimemoved into the partial dryout region and a larger CHF. The reasonwas that the CNT films on the micro-studs made the micro-struc-tured surface more hydrophilic so the liquid could better wet thesurface. The spaces between the carbon nano-tubes also madethe liquid stay longer on the surface with more liquid enteringthe bottom of the grooves to better take advantage of the increasedarea.

The heat transfer also passed through the four regions on boththe micro- and hybrid-structured surfaces as shown in Table 4(pictures in the single phase region were not shown here). Theliquid film began to rupture on the micro-structured surface atheat fluxes of approximately 190 W/cm2 as shown in A2 whilethe hybrid-structured surface was still completely covered withthe liquid film at this heat flux as shown in B2.

The heat fluxes on the 50G � 50S and 100G � 100S micro- andhybrid-structured surfaces are shown in Fig. 7. Adding CNT films tothe 50G � 50S and 100G � 100S micro-structured surfaces had

-5 0 5 10 15 20 25

60

90

120

150

180

210H

eat f

luxe

s (W

/cm

2 )

Tw - T

sat (0C)

25G 25S + SiO2

25G 25S + 4CNT

Fig. 6. Heat fluxes on the 25G � 25S micro- and hybrid-structured surfaces.

Table 4Heated surfaces at various heat fluxes on the 25G � 25S micro- and hy

A. 25G � 25S + SiO2

1

q = 151.2 W/cm2

Tw = 122.0 �C2

q = 190.3 W/cm2

Tw = 115.1 �C3

q = 197.7 W/cm2

Tw = 116.1 �C

34 Z. Zhang et al. / International Journal of Heat and Mass Transfer 80 (2015) 26–37

little effect on the heat transfer due to their groove widths of50 lm and 100 lm, large enough for the droplets to go into thegrooves to use all of the increased area, so the CNT films’ effecton the liquid spread was negligible. Also, the CNT films were onlyon the top surface of the micro-studs, 25% of the base area7.4 mm � 7.4 mm. Therefore, the heat transfer rates on the micro-and hybrid-structured surfaces with the same size micro-studswere similar to each other. The heated surfaces at the various heatfluxes on the 50G � 50S micro- and hybrid-structured surfaces andthe portion where the liquid film ruptures in the partial dryoutregion were similar as shown in Table 5.

Therefore, coating the micro-structured surfaces with carbonnano-tube films having characteristic sizes smaller than the drop-let size more effectively improved the heat transfer than on thesurfaces with larger characteristic sizes.

brid-structured surfaces.

B. 25G � 25S + 4CNT

q = 153.9 W/cm2

Tw = 120.5 �C

q = 191.5 W/cm2

Tw = 119.6 �C

q = 216.9 W/cm2

Tw = 117.8 �C

-5 0 5 10 15 20 25

60

90

120

150

180

210

Hea

t flu

xes

(W/c

m2 )

Tw - T

sat (0C)

50G 50S + SiO2

50G 50S + 4CNT 100G 100S + SiO

2

100G 100S + 4CNT

Fig. 7. Heat fluxes on the 50G � 50S and 100G � 100S micro- and hybrid-structured surfaces.

Table 5Heated surfaces at various heat fluxes on the 50G � 50S micro- an

50G � 50S + + SiO2

1

q = 176.1 W/cm2

Tw = 121.6 �C2

q = 192.0 W/cm2

Tw = 121.0 �C3

q = 205.1 W/cm2

Tw = 122.9 �C

Z. Zhang et al. / International Journal of Heat and Mass Transfer 80 (2015) 26–37 35

3.4. Spray cooling on smooth and hybrid-structured surfaces

The heat fluxes on the smooth and hybrid-structured surfacesare compared in Fig. 8 with the results showing that the heat trans-fer rates were better for the hybrid-structured surfaces, followedby the nano-structured surface and the smooth surface in the thinfilm and partial dryout regions. The heat transfer rates were betteron the 25G � 25S micro-structured surface coated with the CNTfilms than on the 25G � 25S micro-structured surface while addingCNT films on the 50G � 50S and 100G � 100S micro-structuredsurfaces had little effect on the heat transfer as mentioned before.Therefore, the heat transfer was significantly greater on the25G � 25S hybrid-structured surface than on the 50G � 50Shybrid-structured surface with both better than on the100G � 100S hybrid-structured surface. The smooth surface wasalmost dry at the heat flux of only 123.7 W/cm2 as shown in

d hybrid-structured surfaces.

50G � 50S + 4CNT

q = 177.2 W/cm2

Tw = 123.1 �C

q = 193.6 W/cm2

Tw = 121.9 �C

q = 208.2 W/cm2

Tw = 122.5 �C

-5 0 5 10 15 20 25 30

60

90

120

150

180

210q

(W/c

m2 )

Tw - T

sat (0C)

Smooth Smooth + 4CNT 25G 25S + 4CNT 50G 50S + 4CNT 100G 100S + 4CNT

Fig. 8. Heat fluxes on the smooth and hybrid-structured surfaces.

-60 -40 -20 0 200

5

10

15

20

25

30

35

40 Smooth Smooth + SiO

2

Smooth + 1CNT Smooth + 4CNT 25G 25S + SiO

2

50G 50S + SiO2

100G 100S + SiO2

25G 25S + 4CNT 50G 50S + 4CNT 100G 100S + 4CNT

Tm

ax -

Tm

in (

o C)

Tw - T

sat (0C)

0 30 60 90 120 150 180 210 2400

5

10

15

20

25

30

35

40 Smooth Smooth + SiO

2

Smooth + 1CNT Smooth + 4CNT 25G 25S + SiO

2

50G 50S + SiO2

100G 100S + SiO2

25G 25S + 4CNT 50G 50S + 4CNT 100G 100S + 4CNT

Tm

ax -

Tm

in (

o C)

Heat fluxes (W/cm2)

Fig. 9. Maximum and minimum wall temperature differences as a function of thewall superheat and the heat flux.

36 Z. Zhang et al. / International Journal of Heat and Mass Transfer 80 (2015) 26–37

picture A5 in Table 3 while the 25G � 25S hybrid-structured sur-face was still completely covered with a thin liquid film up to aheat flux of approximately 192 W/cm2 as shown in picture B2 inTable 4. The CHF was largest for the 25G � 25S hybrid-structuredsurface with a 75.3% increase over the smooth surface.

3.5. Wall temperature fluctuations

The wall temperatures fluctuated randomly as the spray drop-lets impacted the target surface as discrete drops with the fluctu-ations increased by the unsteady breakup and recovery of theliquid film. The maximum and minimum wall temperature differ-ences are shown in Fig. 9 as functions of the wall superheat andheat flux. The actual upper surface temperature fluctuations mayactually be larger than shown by the extrapolation of themeasurements on the bottom. The maximum and minimum wall

temperature differences first increased with the wall temperatureand the heat flux in the single phase region but then bothdecreased until becoming constant at small values over a wide heatflux range in the thin film region. The fluctuations then increasedrapidly in the partial dryout region until CHF was achieved in thedryout region. Therefore, the wall temperature increases werefairly small (10–20 �C) with small temperature fluctuations (amaximum and minimum difference of approximately 3 �C) as thepower increased over a wide range (90–150 W/cm2).

4. Conclusions

The heat transfer during spray cooling was studied experimen-tally using deionized water to investigate the spray characteristicsand the differences between spray cooling on nano-, micro- andhybrid structured surfaces. The key findings from this study are:

(1) The heat transfer rates were better for the nano-structuredsurface, followed by the smooth surface coated with theSiO2 film and the pure smooth surface since the contactangle was smallest on the nano-structured surface andincreased for the other two surfaces, so the heat transferwas enhanced by higher wettability, although it was notquite significant.

(2) The heat transfer coefficient on the 25G � 25S surface coatedwith the SiO2 film increased much more with the wall tem-perature decreasing more than on the 50G � 50S surfacecoated with the SiO2 film in the partial dryout region dueto the SiO2 film’s stronger hydrophilicity which resulted inmore of the heated area being utilized, while the CHF waslarger on the coated 50G � 50S surface, unlike the heattransfer performance on pure silicon micro-structured sur-faces [11].

(3) Coatings of CNT films on the micro-structured surfaces withcharacteristic sizes smaller than the droplet size more effec-tively improved the heat transfer than on the surfaces withlarger characteristic sizes. The CHF was largest for the25G � 25S hybrid-structured surface with a 75.3% increaseover the smooth surface.

(4) The wall temperature differences and the temperature fluc-tuations were both small for a wide range of heating rates.

Conflict of interest

None declared.

Acknowledgements

This program was supported by the National Basic ResearchProgram of China (Grant No. 2012CB933200) and the Science Fundfor Creative Research Groups of National Natural Science Founda-tion of China (Grant No. 51321002). We thank the National KeyLaboratory of Science and Technology on Micro/Nano Fabricationfor manufacturing the test sections.

References

[1] J. Yang, L.C. Chow, M.R. Pais, Nucleate boiling heat transfer in spray cooling, J.Heat Transfer 118 (3) (1996) 668–671.

[2] R.H. Chen, L.C. Chow, J.E. Navedo, Optimal spray characteristics in water spraycooling, Int. J. Heat Mass Transfer 47 (23) (2004) 5095–5099.

[3] G. Duursma, K. Sefiane, A. Kennedy, Experimental studies of nanofluid dropletsin spray cooling, Heat Transfer Eng. 30 (13) (2009) 1108–1120.

[4] M. Visaria, I. Mudawar, A systematic approach to predicting critical heat fluxfor inclined sprays, J. Electron. Packag. 129 (4) (2007) 452–459.

[5] M. Visaria, I. Mudawar, Effects of high subcooling on two-phase spray coolingand critical heat flux, Int. J. Heat Mass Transfer 51 (21) (2008) 5269–5278.

Z. Zhang et al. / International Journal of Heat and Mass Transfer 80 (2015) 26–37 37

[6] E.A. Silk, J. Kim, K. Kiger, Spray cooling of enhanced surfaces: Impact ofstructured surface geometry and spray axis inclination, Int. J. Heat MassTransfer 49 (25) (2006) 4910–4920.

[7] C.C. Hsieh, S.C. Yao, Evaporative heat transfer characteristics of a water sprayon micro-structured silicon surfaces, Int. J. Heat Mass Transfer 49 (5) (2006)962–974.

[8] Z. Zhang, J. Li, P.X. Jiang, Experiments investigation of spray cooling on flat andenhanced surfaces, Appl. Therm. Eng. 51 (1) (2013) 102–111.

[9] B.H. Yang, H. Wang, X. Zhu, Q. Liao, Y.D. Ding, R. Chen, Heat transferenhancement of spray cooling with ammonia by microcavity surfaces, Appl.Therm. Eng. 50 (1) (2013) 245–250.

[10] H. Bostanci, B.A. Saarloos, D.P. Rini, J.P. Kizito, L.C. Chow, Spray cooling withammonia on micro-structured surfaces, in: 11th Intersociety Conference onThermal and Thermomechanical Phenomena in Electronic Systems, 2008, pp.290–295.

[11] Z. Zhang, P.X. Jiang, X.L. Ouyang, J.N. Chen, D.M. Christopher, Experimentalinvestigation of spray cooling on smooth and micro-structured surfaces, Int. J.Heat Mass Transfer 76 (2014) 366–375.

[12] J.L. Alvarado, Y. Lin, Nanostructured surfaces for enhanced spray cooling, in:27th Army Science Conference, 2010, No. MP-10.

[13] C. Young Lee, M.M. Hossain Bhuiya, K.J. Kim, Pool boiling heat transfer withnano-porous surface, Int. J. Heat Mass Transfer 53 (19) (2010) 4274–4279.

[14] Y. Im, Y. Joshi, C. Dietz, S.S. Lee, Enhanced boiling of a dielectric liquid oncopper nanowire surfaces, Int. J. Micro Nano Scale Transport 1 (1) (2010) 79–96.

[15] S.M. Kwark, M. Amaya, R. Kumar, G. Moreno, S.M. You, Effects of pressure,orientation, and heater size on pool boiling of water with nanocoated heaters,Int. J. Heat Mass Transfer 53 (23) (2010) 5199–5208.

[16] Q. Li, W. Wang, C. Oshman, B. Latour, C. Li, V.M. Bright, Y.C. Lee, R. Yang,Enhanced pool boiling performance on micro-, nano-, and hybrid- structuredsurfaces, in: ASME 2011 International Mechanical Engineering Congress &Exposition, 2011, pp. 633–640.

[17] S. Launay, A.G. Fedorov, Y. Joshi, A. Cao, P.M. Ajayan, Hybrid micro-nanostructured thermal interfaces for pool boiling heat transfer enhancement,Microelectron. J. 37 (11) (2006) 1158–1164.

[18] X.L. Ouyang, P.X. Jiang, R.N. Xu, Thermal boundary conditions of local thermalnon-equilibrium model for convection heat transfer in porous media, Int. J.Heat Mass Transfer 60 (2013) 31–40.

[19] Zhen Zhang, Pei-Xue Jiang, Xiao-Long Ouyang, Jian-Nan Chen, David M.Christopher, Kai-Li Jiang, Experimental investigation of spray cooling on nano-and hybrid-structured surfaces, in: Proceedings of the ASME 2013 4th Micro/Nanoscale Heat & Mass Transfer International Conference, ASME, Hong Kong,China, 2013, pp. V001T01A003.

[20] C. Sodtke, P. Stephan, Spray cooling on micro structured surfaces, Int. J. HeatMass Transfer 50 (19) (2007) 4089–4097.

[21] B. Horacek, K.T. Kiger, J. Kim, Single nozzle spray cooling heat transfermechanisms, Int. J. Heat Mass Transfer 48 (8) (2005) 1425–1438.

[22] M.R. Pais, L.C. Chow, E.T. Mahefkey, Surface roughness and its effects on theheat transfer mechanism in spray cooling, J. Heat Transfer 114 (1) (1992) 211–219.