International Journal of Heat and Mass Transfer 55 (2012) 3849–3856

Contents lists available at SciVerse ScienceDirect

International Journal of Heat and Mass Transfer

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

High heat flux spray cooling with ammonia: Investigationof enhanced surfaces for CHF

Huseyin Bostanci a,⇑, Daniel P. Rini a, John P. Kizito b, Virendra Singh c, Sudipta Seal c, Louis C. Chow c

a RINI Technologies, Inc., 582 S. Econ Circle, Oviedo, FL 32765, United Statesb Department of Mechanical and Chemical Engineering, North Carolina A&T State University, Greensboro, NC 27411, United Statesc Department of Mechanical, Materials and Aerospace Engineering, University of Central Florida, Orlando, FL 32816, United States

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

Article history:Received 2 June 2011Received in revised form 5 March 2012Available online 20 April 2012

Keywords:Spray coolingThermal managementCHFEnhanced surfaces

0017-9310/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.ijheatmasstransfer.2012.03

⇑ Corresponding author.E-mail address: huseyin.bostanci@rinitech.com (H

A spray cooling study was conducted to investigate the effect of enhanced surfaces on Critical Heat Flux(CHF). Test surfaces involved micro-scale indentations and protrusions, macro (mm) scale pyramidal pinfins, and multi-scale structured surfaces, combining macro and micro-scale structures, along with asmooth surface that served as reference. Tests were conducted in a closed loop system using a vaporatomized spray nozzle with ammonia as the working fluid. Nominal flow rates were 1.6 ml/cm2 s ofliquid and 13.8 ml/cm2 s of vapor, resulting in a pressure drop of 48 kPa. Results indicated that themulti-scale structured surface helped increase maximum heat flux limit by 18% over the referencesmooth surface, to 910 W/cm2 at nominal flow rate. During the additional CHF testing at higher flowrates, most heaters experienced failures before reaching CHF at heat fluxes above 950 W/cm2. However,some enhanced surfaces can achieve CHF values of up to �1100 W/cm2 with �67% spray cooling effi-ciency based on liquid usage. The results also shed some light on the current understanding of the spraycooling heat transfer mechanisms. Enhanced surfaces are found to be capable of retaining more liquidcompared to a smooth surface, and efficiently spread the liquid film via capillary force within the struc-tures. This important advantage delays the occurrence of dry patches at high heat fluxes, and leads tohigher CHF. The present work demonstrated ammonia spray cooling as a unique alternative for challeng-ing thermal management tasks that call for high heat flux removal while maintaining a low device tem-perature with a compact and efficient cooling scheme.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction ment in heat transfer coefficient with decreasing surface roughness

Many critical applications today, in electronics, optics and aero-space fields, among others, demand advanced thermal manage-ment solutions for the acquisition of high heat loads theygenerate in order to operate reliably and efficiently. Current com-peting technologies for this challenging task include several singleand two phase cooling options. When these cooling schemes arecompared based on the high heat flux removal (100–1000 W/cm2) and isothermal operation (within several �C across the cooleddevice) aspects, as well as system mass, volume and power con-sumption, spray cooling appears to be the best choice.

Although a vast amount of research has been done on heat trans-fer enhancement in general, studies focusing on spray coolingenhancement are fairly limited. In early works, Pais et al. [1] andSehmbey et al. [2] examined the effects of surface roughness andcontact angle using water with air atomized nozzle, at flow ratesup to 1.4 ml/cm2 s water and 400 ml/cm2 s air, and found enhance-

ll rights reserved..040

. Bostanci).

and increasing contact angle. They obtained heat fluxes up to1250 W/cm2 at 11 �C surface superheat on ultrasmooth (Ra = 0.3lm) copper surface.

Kim et al. [3] investigated spray cooling enhancement on micro-porous coated surfaces using water at flow rates up to 0.03 ml/cm2 s. The porous layer was fabricated using a mixture ofmethyl–ethyl-ketone (MEK), epoxy, and aluminum powder, andits maximum thickness was 500 lm. They found that the CHFincreased 50% relative to the uncoated surface. However, highestheat flux reached was 3.2 W/cm2 at 65 �C surface superheat dueto very low flow rates.

Stodke and Stephan [4] studied spray cooling on microstruc-tured and micro-porous surfaces using water at 6 kPa system pres-sure and 1.4 ml/cm2 s flow rate. Pyramidal micro-grooves andmicro-pyramids with 75 lm height increased the wetted area bya factor of 1.4. Their 100-lm thick porous layers were very similarto those used by Kim et al. [3] and created with the same ingredi-ents of MEK, epoxy, and aluminum powder. A maximum heat fluxof 97 W/cm2 was observed for the micro-pyramid surface com-pared to 30 W/cm2 for the flat surface, both at a surface superheatof 12 �C. This enhancement was much larger than the surface area

Nomenclature

A area (cm2)CHF Critical Heat Flux (W/cm2)CHFref Critical Heat Flux of reference surface (W/cm2)cp liquid specific heat (J/kg �C)d32 Sauter-mean droplet size (m)EFCHF CHF enhancement factorh heat transfer coefficient (W/m2 �C)hfg latent heat of vaporization (kJ/kg)I current (A)k heater wall thermal conductivity (W/m �C)N mean droplet flux (#/m2 s)q00 heat flux (W/cm2)Tl liquid temperature (�C)Tsat saturation temperature (�C)

Tsurf surface temperature (�C)TCavg average thermocouple reading (�C)Ra average surface roughness (lm)V voltage (V)v mean droplet velocity (m/s)_V liquid flow rate of spray nozzle (ml/cm2 s)x vertical distance between TC hole to spray surface in

heater wall (m)DTsat surface superheat (�C)DTsub subcooling (�C)e spray cooling effectiveness (J/ml)ql liquid density (kg/m3)g spray cooling efficiency

3850 H. Bostanci et al. / International Journal of Heat and Mass Transfer 55 (2012) 3849–3856

enhancement. However, when a micro-porous surface was used, asignificant degradation in heat transfer occurred when comparedto the plain surface resulting in 14 W/cm2 maximum heat flux at12 �C superheat. The low heat flux was due to the poor thermalconductivity of the epoxy binder.

Amon et al. [5] and Hsieh and Yao [6] performed spray coolingexperiments with water on silicon substrates using 1 � 2 nozzlearray at very low flow rates of up to 0.07 ml/cm2 s on square mi-cro-studs with 160–480 lm heights. The surface texture was foundto have little effect in the single phase and dry out regimes. Theauthors attributed the higher heat transfer observed for the mi-cro-textured surfaces in the intermediate regimes to more effectivespreading of the liquid by capillary forces. The maximum heat fluxachieved was just over 50 W/cm2 at 55 �C surface superheat.

Silk et al. [7] investigated the effect of surface geometry in spraycooling with PF-5060 using 2 � 2 nozzle array at 1.6 ml/cm2 s flowrate. They used embedded structures (dimples, pores and tunnels)and compound extended structures (straight fins, cubic pin finsand dimples) all in the order of 1 mm size. Of these macrostruc-tured surfaces, straight fins and porous tunnels performed the bestproviding a CHF of 175 W/cm2 for gassy conditions at surfacesuperheats of up to 36 �C and offered CHF enhancement of 62%over flat surface.

Coursey et al. [8] performed spray cooling tests focusing onstraight fin geometry with heights between 0.25 and 5 mm. Theyused PF-5060 at flow rates up to 1.0 ml/cm2 s and found that finheights between 1 and 3 mm were optimum for heat fluxes up to124 W/cm2 at 19 �C surface superheat.

In the earlier work, Bostanci et al. [9] has evaluated two kinds ofmicrostructured surfaces featuring indentations (micro-i) and pro-trusions (micro-p), in spray cooling tests with saturated anhydrousammonia. They used vapor atomized spray nozzles with 1.6 ml/cm2 s liquid and 13.8 ml/cm2 s vapor flow rates. At heat flux levelsof up to 500 W/cm2, they observed 49% and 112% improvement inheat transfer coefficient with micro-i and micro-p surfaces, respec-tively, over a reference smooth surface.

The current study also focused on high heat flux spray coolingwith ammonia on enhanced surfaces. Compared to some othercommonly used coolants, ammonia possesses important advanta-ges such as low saturation temperature, and a relatively high latentheat of vaporization. Moreover, enhanced surfaces offer a potentialto greatly improve heat transfer performance. The main objectivesof the study were to investigate the effect of surface enhancementon spray cooling Critical Heat Flux (CHF) limit, and contribute tothe current understanding of spray cooling heat transfer mecha-nisms. The experimental study used a set of optimized enhancedsurfaces, as detailed by Bostanci [10], including micro-scale inden-tations and protrusions, macro (mm) scale pyramidal pin fins, and

multi-scale structured surfaces, combining macro and micro-scalestructures.

2. Experimental setup and procedure

Experiments were conducted in a closed loop spray cooling sys-tem. Fig. 1 is a schematic diagram of the system where the maincomponents consist of a reservoir, 1 � 2 nozzle array, a subcooler,a condenser, and a pump. In the setup, the reservoir suppliesammonia liquid and vapor to the nozzle array. Liquid and vapormix in the atomizer nozzles and the resulting sprays cool a1 cm � 2 cm heater where a mounted thick film resistor is the heatsource. Exhaust from 1 � 2 nozzle array slightly subcools theincoming liquid supply in a small heat exchanger before flowinginto the larger heat exchanger to condense. Finally, the two phasepump takes the liquid and vapor ammonia and transfers it back tothe reservoir, providing the pressure difference that is needed todrive ammonia in the cycle and generate the spray. A separateair cooled R-22 cycle is employed to absorb the heat from theammonia cycle and reject it to the ambient. System allows control-ling flow rates and pressures across the nozzle array and isequipped with computer controlled data acquisition system foraccurate data recording.

2.1. Working fluid

In the present work, anhydrous ammonia (Refrigerant 717) wasused as the working fluid. Ammonia has the second highest latentheat of vaporization after water among refrigerants (1368 and2257 kJ/kg at atmospheric pressure, respectively). For applicationsthat require low temperature operation, ammonia becomes advan-tageous compared to water by offering lower saturation tempera-ture at a given pressure. Ammonia however, is not compatible withmost of the commonly used engineering materials. The use ofammonia requires careful material selection and component de-sign in cooling systems.

2.2. Spray nozzle

RINI’s highly compact 1 � 2 vapor atomized nozzle array,shown in Fig. 2a, was used in all tests. In these nozzles a fine liquidstream is injected into a high velocity vapor stream. The shearforce created by the vapor stream atomizes the liquid into finedroplets that are ejected from the nozzle orifice. The spray nozzlesfeatured a 0.76 mm orifice size, and for the flow range of 1.5–2.5 ml/cm2 s liquid and 12.0–20.0 ml/cm2 s vapor, they provideda full-cone spray pattern. Fig. 2a also shows the pressure andtemperature measurement ports on the nozzle array that were

Fig. 1. Schematic diagram of the spray cooling setup.

Exhaust outlet

Vapor-atomized nozzles

Liquid inlet

1 cm

Pressure / temperature measurement ports

Vapor inlet

Thick film resistor

TC#2

TC#1

a

b

c dFig. 2. RINI 1 � 2 vapor atomized nozzle array (a), heater-spray side (b), heater-resistor side (c), and complete 1 � 2 nozzle array and heater assembly (d).

H. Bostanci et al. / International Journal of Heat and Mass Transfer 55 (2012) 3849–3856 3851

needed to determine the driving pressures across the nozzle, andsaturation condition of the coolant. Nozzle-to-surface distance of11 mm was found to be adequate for each spray to cover 1 cm2

area, and the distance was held constant.

2.3. Heaters and spray surfaces

The heater design employed in this study is shown in Fig. 2band c. Heater body and enhanced structures were made of coppermaterial, and featured a nickel coated spray surface for corrosionprotection. The current study is the continuation of a previous ef-

fort that focused on investigation of the highest heat transfer coef-ficient on enhanced surfaces using the same experimental setup,working fluid and nozzle array. Heater design however, was mod-ified to accommodate the higher heat loads required in the presentCHF tests. Total heated area on the current design was reduced to1 cm � 1 cm featuring a 1 cm2 thick film resistor that was solderedonto heater to simulate heat generating device. Heat flux wasdetermined from the total power supplied into the thick film resis-tor per unit resistor base area ðq00 ¼ V � I=AÞ. Heater temperaturewas monitored with two type-T thermocouples embedded halfwayin the heater wall and spaced equally across the 1 cm2 area as

Table 1Heater descriptions.

Heater ID Surface conditionEnhancement type, structure geometry, structure size

s1 Smooth, plain, machine-finished (Ra � 0.3 lm)s2 Smooth, plain, machine-finished (Ra � 0.3 lm)mi-c Microstructured, indentations, coarse (Ra � 5 lm)mp-c Micro-structured, protrusions, coarse (Ra � 20 lm)Mpf-0.50 Macro structured, pyramidal fins, 0.50 mmMpf-0.50mi-c Multi-scale structured, pyramidal fins/indentations,

0.50 mm/coarse (Ra � 5 lm)Mpf-0.50mp-c Multi-scale structured, pyramidal fins/protrusions,

0.50 mm/coarse (Ra � 20 lm)

3852 H. Bostanci et al. / International Journal of Heat and Mass Transfer 55 (2012) 3849–3856

illustrated in Fig. 2c. Temperature spread resulted from the twothermocouples did not exceed ±1 �C at 200 W/cm2, and ±1.5 �C at500 W/cm2 indicating uniform cooling across the surface. Spraysurface temperature was calculated by extrapolating average oftwo thermocouple readings through the known distance to thesurface and assuming steady 1-D conduction through the heaterwall ðTsurf ¼ TCavg � ðq00 � xÞ=kÞ.

The description and surface characteristics of the test heatersinvestigated in this study are provided in Table 1. Two of the heat-ers, with machine finished-smooth surface, served as reference forall enhanced surfaces in obtaining quantitative performance com-parison. Enhanced surfaces, that can be grouped as micro-, macro-and multi-scale structured surfaces, reflected the optimized struc-ture geometry and size based on a study by Bostanci [10].

Fig. 3 includes some schematics, scanning electron microscope(SEM) images, and solid models of micro- and macrostructuredsurfaces to characterize the surface structures. The microstruc-tured spray surfaces featuring indentations and protrusions werefabricated using a particle blasting and a thermal spray coating

Fig. 3. Schematics and SEM images of microstructured surfa

process, respectively. While the smooth surface had a Ra roughnesslevel of 0.3–0.5 lm, the surfaces with indentations and protrusionsresulted in Ra of �5 lm and 20 lm. Although not obvious from theSEM images, the surface mi has open cavities, while the surface mphas many randomly sized re-entrant cavities. The macrostructuredsurface featured 0.50 mm high pyramidal fins (with 0.59 mm baselength and 0.59 mm pitch) that were fabricated on a precise CNCmill.

2.4. Test conditions and procedure

All the tests were conducted at 550–570 kPa (65–68 psig) and7–8 �C conditions using saturated ammonia as the working fluid.Cooling curves that were in the form of surface superheat(DTsat = Tsurf � Tsat) vs. heat flux (q00) were generated for each testheater by increasing the heat flux gradually and recording the cor-responding heater temperatures at certain sampling rates.

A series of tests were performed in order to find the optimumflow rates that balances high heat transfer rate, and reasonablylow coolant usage and pumping power, as detailed in [10]. In thesetests, liquid flow rate was first varied between 1.4 and 1.8 ml/cm2 s, at a constant vapor flow rate of 15.7 ml/cm2 s. Data illus-trated that the increase in liquid flow rate up to 1.6 ml/cm2 s im-proves the heat transfer, but further increase results inpractically the same cooling curve. In the next step, effect of vaporflow rate was examined for the constant 1.6 ml/cm2 s liquid flowrate. Data then suggested that varying vapor flow rate in the rangeof 11.8–15.7 ml/cm2 s have no considerable effect on the heattransfer. Based on these data, optimum flow rates were determinedto be 1.6 ml/cm2 s for liquid, and 13.8 ml/cm2 s for vapor (corre-sponding to a pressure drop of 48 kPa (7 psi) across the nozzle),and these conditions were applied in the initial tests to investigateCHF performance of enhanced surfaces. However, higher flow rates

ces (a), and solid model of macrostructured surface (b).

100

200

300

400

500

600

700

800

900

1000

q" (W

/cm

2 )

s1s2Ref

CHF

Nominal (Low) Flow Rate (1.6 ml/cm2.s liquid, 13.8 ml/cm2.s vapor)

H. Bostanci et al. / International Journal of Heat and Mass Transfer 55 (2012) 3849–3856 3853

of up to 2.1 ml/cm2 s for liquid, and 17.7 ml/cm2 s for vapor (corre-sponding to a pressure drop of 83 kPa (12 psi) across the nozzle),were employed as well to determine the effect of flow rate onCHF limits.

Additional tests were performed to determine whether themagnitude and duration of the heat flux steps in the generationof cooling curves were a factor. Comparison of data from thesetests matched very well and indicated flexibility in the selectionof the heat flux steps in tests. Heat flux was therefore gradually in-creased in steps of 10–100 W/cm2 up to CHF and correspondingheater temperatures are recorded every 1–3 s over 1–2 min longsteps. Adjusting steps to be smaller and quicker for the high heatflux ranges enabled keeping the overall testing time at a reasonablelevel while approaching CHF in a smooth and controllable pace.

00 10 20 30 40 50 60

ΔTsat (oC)

Fig. 4. CHF performance of the reference surface based on two different heaters.

0

100

200

300

400

500

600

700

800

900

1000

q"

(W/c

m2 )

Ref, EF_CHF=1.00

mi-c, EF_CHF=1.01

mp-c, EF_CHF=1.18

Mpf-0.50, EF_CHF=1.10

Mpf-0.50mi-c, EF_CHF=1.06

Mpf-0.50mp-c, EF_CHF=1.18

CHF

Nominal (Low) Flow Rate (1.6 ml/cm2.s liquid, 13.8 ml/cm2.s vapor)

2.5. Uncertainty analysis

Uncertainties were estimated mainly for heat flux and temper-ature measurements that are critical in performance evaluation.Error involved in heat flux measurement (considering variationsin voltage, current, and area) was ±2.2% at 750 W/cm2. Error intemperature measurements from the embedded thermocouplesin the heater wall was calibrated to be ±0.2 �C. Spray surface tem-perature had an uncertainty of approximately ±1.0 �C involvinguncertainty in temperature extrapolation from the TC hole to theheater surface at 750 W/cm2. Heat transfer coefficients included±3.5, 4.1 and 4.4% uncertainty at 750 W/cm2 and nominal flowrates for the surfacess, mp-c, and Mpf-0.50, respectively. Flow ratemeasurements had ±5% uncertainty dictated by the flow metercharacteristics. Heat loss from the thick film resistor to surround-ing heater body via conduction was estimated to be approximately9% based on finite element analysis results. Heat loss to the ambi-ent environment was negligibly small (<1 W) based on calculationsconsidering natural convection and black body radiation from150 �C heater surface to 20 �C stagnant air.

0 10 20 30 40 50 60

ΔTsat (oC)

Fig. 5. CHF performance of the reference, and micro-, macro-, and multi-scalestructured surfaces.

3. Results and discussion

Two heaters with the surface s were tested to obtain referencedata. As shown in Fig. 4, the difference in performance betweenthese two surfaces is negligible for the entire testing range. Oneof the heaters reached CHF at 760 W/cm2, while the other one at-tained a higher CHF at 780 W/cm2. A reference cooling curve wasthen established by averaging the two data sets.

Study was continued by testing all the enhanced surfaces atfixed liquid/vapor flow rates to evaluate their CHF performance.Results are included in Fig. 5 along with the reference curve forcomparison. Among two types of microstructured surfaces, thesurface mi-c reached CHF at 780 W/cm2, while the surface mp-c reached CHF at a much higher level at 910 W/cm2. These sur-faces also exhibited different cooling curves. The surface mp-centered two-phase regime at lower heat fluxes, and had higherheat transfer coefficients up to 500 W/cm2. At higher heat fluxeshowever, the surface mi-c performed better and reached CHFsooner. The surface mp-c transitioned to the last region of cool-ing curve much slower resulting in a higher CHF and surfacesuperheat.

The macrostructured surface Mpf-0.50, on the other handreached CHF at 850 W/cm2, and indicated a superior performancethroughout the testing range compared to the surface s.

When it comes to multi-scale structured surfaces that combinemicro- and macro-scale structures, the surface Mpf-0.50mi-c, per-formed better than the surface mi-c, and reached CHF at 820 W/cm2. The other surface, Mpf-0.50mp-c, performed same as the sur-face mp-c, and provided a CHF of 910 W/cm2.

As far as the overall heat transfer performance is concerned, thesurface mp-c was better than the surface mi-c up to 500 W/cm2,where two curves crossed over, and at higher heat fluxes the sur-face mi-c offered higher heat transfer coefficients. The contributionof 0.50 mm high pyramids to the performance of microstructures(indentations and protrusions) was not significant as evidencedby �1.5 �C lower superheat at 500 W/cm2.

Fig. 5 also lists an enhancement factor, EFCHF, that is defined as

EFCHF ¼CHF

CHFrefð1Þ

to express each heater’s CHF enhancement over the reference sur-face. As can be noticed, while the surface mi-c offers a minimal1% improvement, the surface mp-c provides 18% increase in CHFover the surface s. The surface Mpf-0.50 alone offers 10% enhance-ment. When the surface Mpf-0.50 is combined with the surfacesmi-c and mp-c, they result in 6% and 18% improvement, respec-tively. These data therefore suggest that CHF enhancement due tothe multi-scale structures is not additive.

Current results can provide some insights into spray coolingCHF enhancement mechanism. In general, rough, porous or tex-tured surfaces would retain more liquid compared to a smoothsurface for a given spray nozzle and flow rate. The textured sur-faces also provide an efficient means to spread the liquid film via

0

100,000

200,000

300,000

400,000

0 100 200 300 400 500 600 700 800 900 1000

q" (W/cm2)

h (W

/m2o

C)

Refmi-cmp-cMpf-0.50Mpf-0.50mi-cMpf-0.50mp-c

Nominal (Low) Flow Rate (1.6 ml/cm2.s liquid, 13.8 ml/cm2.s vapor)

Fig. 6. Heat transfer coefficients as a function of heat flux for reference, and micro-,macro-, and multi-scale structured surfaces.

Table 2Flow rates used in CHF tests.

Condition Liquid flow rate(ml/cm2 s)

Vapor flow rate(ml/cm2 s)

Liquid-to-vaporloading ratioa

Nominal(low)

1.6 13.8 14.7

Medium 1.8 15.7 15.0High 2.1 17.7 15.2

a Based on mass flow rate.

3854 H. Bostanci et al. / International Journal of Heat and Mass Transfer 55 (2012) 3849–3856

capillary force within the micro-scale structures. These observa-tions have also been reported by several experimental studies[1,3,6]. The capillary wicking delays the occurrence of dry patchesat high heat fluxes and leads to higher CHF.

In the current study, based on both qualitative visual observa-tions, and quantitative surface roughness analysis results as de-tailed in [10], the surface mp-c has the highest roughness, overallstructure height, and actual surface area with plenty of re-entrant

123456789

1011

q" (W

/cm

2 )

0100200300400500600700800900

10001100

0 10 20 30 40 50 6ΔTsat (oC)

q" (W

/cm

2 )

1s, low1s, medium1s, high

CHF

123456789

1011

q" (W

/cm

2 )

0100200300400500600700800900

10001100

0 10 20 30 40 50 6ΔTsat (oC)

q" (W

/cm

2 )

Mpf-0.50, lowMpf-0.50, high

CHF Heater Limit

Fig. 7. CHF performance of the reference, and micro-, macro-, and

cavities. Therefore the surface mp-c is expected to hold more li-quid, and the multitude of cavities can spread the liquid very effi-ciently, thus keeping the surface wet longer and achieving higherCHF values. The surface mi-c on the other hand, exhibits lessroughness and open cavities, that can still hold more liquid thana smooth surface, but cannot resist liquid film break up as effi-ciently as the surface mp-c at high fluxes leading to only a slightCHF improvement over the smooth surface. The surface Mpf-0.50naturally forms grooves between adjacent pyramids that can helpmanage the liquid distribution. However, as experimental data im-plied, its enhancement level is between the two aforementionedmicrostructured surfaces.

The fluid retention and capillary wicking mechanism also ex-plains why the surface mp-c performs better than the surfacemi-c up to a certain heat flux, but the trend reverses afterwardsduring CHF tests. Although surface bubble nucleation is very effec-tive at low to medium heat fluxes (<500 W/cm2), evaporative ef-fects gradually become more important as the controlling factorat higher heat fluxes. Hence, the surface mi-c with thinner liquidfilm starts to offer higher heat transfer coefficients beyond500 W/cm2. Eventually the thin film breaks up resulting in morepronounced dry patches, and the heater approaches CHF sooner.The surface mp-c, with a thicker liquid film, results in a lowerevaporation rate and consequently lower heat transfer coefficients,but also extends the transition to CHF.

Fig. 6 includes heat transfer coefficients as a function of heatflux. The curves peak at 500–700 W/cm2 heat flux, and then pres-ent a declining slope as they approach to CHF. The highest heattransfer coefficient with the surface s was 220,000 W/m2�C. Amongothers, surfaces mi-c, Mpf-0.50mi-c, and Mpf-0.50mp-c all reachedheat transfer coefficient of approximately 300,000 W/m2�C.

Once the CHF limits of the enhanced surfaces were determinedusing the nominal flow rates of 1.6 ml/cm2 s liquid, and 13.8 ml/cm2 s vapor, the study was continued to investigate the effect offlow rate on CHF. For spray cooling, it has been established thatincreasing the liquid flow rate would help increase CHF for a givennozzle. However, the increase in CHF is not proportional to theflow rate, and the higher flow rate is only effective up to a certainlevel, beyond which CHF remains relatively the same regardless of

00000000000000000000000

0 10 20 30 40 50 60ΔTsat (oC)

mi-c, lowmi-c, highmp-c, lowmp-c, high

CHF Heater Limit

0

00000000000000000000000

0 10 20 30 40 50 60ΔTsat (oC)

Mpf-0.50mi-c, lowMpf-0.50mi-c, highMpf-0.50mp-c, lowMpf-0.50mp-c, high

CHF Heater Limit

0

multi-scale structured surfaces at various flow rate conditions.

650

700

750

800

850

900

950

1000

1050

1100

1150

1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2

Liquid Flow Rate (ml/cm2.s)

CHF

(W/c

m2 )

Refmi-cmp-cMpf-0.50Mpf-0.50mi-cMpf-0.50mp-c

Heater limit

Fig. 8. Effect of liquid flow rate on CHF for the reference, and micro-, macro-, andmulti-scale structured surfaces.

40

50

60

70

80

90

1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2Liquid Flow Rate (ml/cm2.s)

ηat

CHF

(%)

308

385

462

539

616

693

εat

CHF

(J/m

l)

Refmi-cmp-cMpf-0.50Mpf-0.50mi-cMpf-0.50mp-c

η and ε at heater limit

Fig. 9. Effect of liquid flow rate on spray cooling efficiency and effectiveness for thereference, and micro-, macro-, and multi-scale structured surfaces.

H. Bostanci et al. / International Journal of Heat and Mass Transfer 55 (2012) 3849–3856 3855

flow rate [11,12]. This can be attributed to the counterbalance ofvarious spray cooling heat transfer mechanisms driven by theadvantage of higher droplet velocity and the disadvantage of high-er film thickness. Other studies [13,14] use the mean droplet veloc-ity (v), the mean spray droplet flux (N), and the Sauter-meandroplet diameter (d32) as three independent spray parameterswhich affect the CHF. These studies, utilizing extensive experimen-tal data, found that CHF varies with V1/4 and N1/6, and is relativelyindependent of d32. Although these spray parameters are not spe-cifically measured in the current work, the parameters are propor-tional to flow rate, where an increase in flow rate would increase V,N and d32 simultaneously.

The use of higher flow rates for improved CHF performance alsobrings system level implications, such as higher pumping power,affects cooling efficiency, and thus requires further optimization.When CHF is the main design consideration, however, higher flowrates would be preferred. The present work therefore investigatedthe CHF limits of the enhanced surfaces at the flow rates listed inTable 2. Fig. 7 shows all the data from CHF tests at low, medium,and high flow rates. The figure includes four separate plots for eas-ier comparison between reference, and micro-, macro-, and multi-scale structured surfaces. The surface s was tested at all three flowrate conditions, and the CHF values at medium and high flow rateswere the same at 930 W/cm2. Data thus suggested that increasingflow rate beyond the medium level has no considerable effect onCHF.

Once this trend was established, the other surfaces were testedonly at high flow rates in addition to nominal/low flow rates. Allheaters with enhanced surfaces, however, consistently failed dur-ing high flow rate CHF tests at heat fluxes starting at approximately960 W/cm2. These data points are marked as ‘‘heater limit’’ in theplots, to distinguish them from the CHF condition. Consideringthe elevated temperatures at these heat flux levels, cracking ofthe thick film resistors was most likely due to the stresses inducedby thermal expansion mismatch. As a result, true CHF value forthese conditions could not be experimentally obtained. The sur-faces Mpf-0.50mp-c and mp-c attained the highest heat flux, butnot CHF, of 1090 W/cm2 before the heater damage occurred.

As far as the overall heat transfer performance, besides CHF, isconcerned, higher flow rates only aided the surface s where higherheat transfer coefficients were achieved. For other surfaces, highflow rate generally resulted in slightly higher superheats at heatfluxes lower than 700 W/cm2.

In an effort to further confirm the effect of flow rate on CHF,some additional tests were conducted with the surface s at med-ium and high flow rates. Data indicated that CHF varies between760 and 780 W/cm2 for low, 890–930 W/cm2 for medium, and920–930 W/cm2 for the high flow rates, and suggest that CHF isnot necessarily a very repeatable quantity, and can occur over anarrow heat flux range. Based on this set of data in Fig. 8, a curvewas fitted to represent the effect of flow rate on the CHF for theoutlined conditions. The trend once again implies that increasingflow rate beyond a certain level (the medium flow rate in this case)has a minimal effect on CHF, and is consistent with observationsmade by earlier studies. Fig. 8 also incorporates CHF data, or high-est recorded heat fluxes at heater limit otherwise, from all othertests. Using the curve fit for the surface s as a guideline, it was off-set to estimate the effect of flow rate on CHF for other surfaces. Theamount of offset was determined based on the experimental CHFvalues at low flow rate. During a close examination of the data, thisapproach actually seems reasonable since heat fluxes at heater lim-it, for the surfaces Mpf-0.50 and Mpf-0.50mp-c at high flow rate,match or exceed the estimated CHF values.

As mentioned before, when attempting to achieve higher CHFwith higher flow rates, another performance aspect to consider isthe spray cooling efficiency. In general it can be defined as the ratio

of the actual heat removed to the total heat capacity of the liquidused, including the required heat to bring the liquid from sub-cooled to saturation condition (sensible heat), and then to com-plete vaporization (latent heat). This liquid usage efficiency g,can be expressed as

g ¼ q00

_VqlðcpDTsub þ hfgÞð2Þ

where _V is volumetric flow rate, ql is liquid density, cp is liquid spe-cific heat, DTsub = Tsat � Tl is subcooling, and hfg is latent heat ofvaporization. Since saturated spray conditions were maintainedthroughout this study, total heat capacity of the used liquid isequivalent to its latent heat of vaporization. Spray cooling efficien-cies of all surfaces at their respective CHF values (or heater limits)are summarized in Fig. 9 as a function of liquid flow rate. At lowflow rate condition, efficiency can be seen to range from 63.4%,for the surface s, to 74.9% for the surfaces mp-c and Mpf-0.50mp-c. Complete data set from the reference surface was again used todetermine the overall trend, which indicates that when flow rateis increased from low to medium, efficiency slightly goes up(�1%), but with further flow rate increase efficiency starts to de-crease since CHF remains nearly the same. The highest efficiencyreached during this study was therefore approximately 75% for boththe surfaces mp-c and Mpf-0.50mp-c.

Besides efficiency, another performance parameter, spray cool-ing effectiveness e, with the resulting units of J/ml is defined as

e ¼ q00A_V

ð3Þ

3856 H. Bostanci et al. / International Journal of Heat and Mass Transfer 55 (2012) 3849–3856

and can be useful in comparison of alternative coolants and spraynozzle designs. Effectiveness values from all CHF tests are also in-cluded in Fig. 9.

4. Conclusions

The present work focused on high heat flux spray cooling withammonia. A set of optimized enhanced surfaces, including micro-scale indentations and protrusions, macro (mm) scale pyramidalpin fins, and multi-scale structured surfaces, combining macroand micro-scale structures, was investigated to determine theCHF limits and gain a better understanding on the underlyingmechanism. Based on the results, the following conclusions aredrawn:

� The surfaces Mpf-0.50mp-c and mp-c had the highest CHF valueof 910 W/cm2 with the nominal flow rates of 1.6 ml/cm2 s liquidand 13.8 ml/cm2 s vapor, corresponding to 18% increase overthe reference smooth surface.� When the effect of higher liquid flow rate was investigated,

most of the heaters experienced resistor failures at heat fluxes>950 W/cm2 before they reach CHF. However, the effect of flowrate was still captured, and it was estimated that the surfaceMpf-0.50mp-c can attain a CHF value of approximately1100 W/cm2.� Enhanced surfaces are capable of retaining more liquid com-

pared to a smooth surface, and efficiently spread the liquid filmvia capillary force within the structures. Fluid retention delaysthe occurrence of dry patches at high heat fluxes, and leads tohigher CHF.� Overall, the present study demonstrated that the spray cooling

with ammonia on enhanced surfaces offers a very high level ofheat flux removing capability, which previously considered tobe achievable by using water only. Having ammonia as a highlycapable coolant alternative would enable unique thermal man-agement applications that demand operation of high powerdevices at low temperatures, such as laser systems.

Acknowledgments

We acknowledge US Air Force Research Laboratory (AFRL) Pro-pulsion Directorate and Universal Technology Corporation for theirfinancial support.

References

[1] 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 (1992) 211–219.

[2] M.S. Sehmbey, M.R. Pais, L.C. Chow, Effect of surface material properties andsurface characteristics in evaporative spray cooling, J. Thermophys. HeatTransfer 6 (1992) 505–511.

[3] J.H. Kim, S.M. You, U.S. Choi, Evaporative spray cooling of plain andmicroporous coated surfaces, Int. J. Heat Mass Transfer 47 (2004) 3307–3315.

[4] C. Stodke, P. Stephan, Spray cooling heat transfer on microstructured surfaces,in: Proceedings of the 6th World Conference on Experimental Heat Transfer,Fluid Mechanics, and Thermodynamics, Matsushima, Miyagi, Japan, 2005.

[5] C. Amon, S.C. Yao, C.F. Wu, C.C. Hsieh, Microelectromechanical system-basedevaporative thermal management of high heat flux electronics, J. Heat Transfer127 (2005) 66–75.

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

[7] E. Silk, J. Kim, K.T. Kiger, Enhanced surface spray cooling with embedded andcompound extended surface structures, in: Proceedings of ITHERM 2006, SanDiego, CA, 2006.

[8] J. Coursey, J. Kim, K.T. Kiger, Spray cooling of high aspect ratio openmicrochannels, in: Proceedings of ITHERM 2006, San Diego, CA, 2006.

[9] H. Bostanci, D.P. Rini, J.P. Kizito, L.C. Chow, Spray cooling with ammonia onmicrostructured surfaces: performance enhancement and hysteresis effect, J.Heat Transfer 131 (2009). 071401-1-9.

[10] H. Bostanci, High Heat Flux Spray Cooling with Ammonia on EnhancedSurfaces, Ph.D. Dissertation, University of Central Florida, Orlando, Florida,2010.

[11] L.C. Chow, M.S. Sehmbey, M.R. Pais, High heat flux spray cooling, in: C.-L. Tien(Ed.), Annual Review of Heat Transfer, vol. 8, Hemisphere Pub. Corp., 1997, pp.291–318.

[12] J. Yang, M.R. Pais, L.C. Chow, Critical heat flux limits in secondary gas atomizedliquid spray cooling, Exp. Heat Transfer 6 (1993) 55–67.

[13] R.-H. Chen, L.C. Chow, J.E. Navedo, Effects of spray characteristics on critical insubcooled water spray cooling, Int. J. Heat Mass Transfer 45 (2002) 4033–4043.

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