International Journal of Heat and Fluid Flow 42 (2013) 176–189

Contents lists available at SciVerse ScienceDirect

International Journal of Heat and Fluid Flow

journal homepage: www.elsevier .com/ locate / i jhf f

Two-phase flow operational maps for multi-microchannel evaporators

0142-727X/$ - see front matter � 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.ijheatfluidflow.2013.03.006

⇑ Corresponding author. Tel.: +41 21 693 59 81; fax: +41 21 693 59 60.E-mail address: john.thome@epfl.ch (J.R. Thome).URL: http://www.ltcm.epfl.ch (J.R. Thome).

Sylwia Szczukiewicz, Navid Borhani, John Richard Thome ⇑Laboratory of Heat and Mass Transfer (LTCM), École Polytechnique Fédérale de Lausanne (EPFL), Station 9, CH-1015 Lausanne, Switzerland

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

Article history:Received 11 July 2012Received in revised form 5 March 2013Accepted 12 March 2013Available online 18 April 2013

Keywords:Two-phase refrigerant flowSilicon multi-microchannel evaporatorInfra-red thermographyFlow visualizationOperating regimeOperational map

The current paper presents new operational maps for several different multi-microchannel evaporators,with and without any inlet restrictions (micro-orifices), for the two-phase flow of refrigerants R245fa,R236fa, and R1234ze(E). The test fluids flowed in 67 parallel channels, each having a cross-sectional areaof 100 � 100 lm2. In order to emulate the power dissipated by active components in a 3D CMOS CPUchip, two aluminium microheaters were sputtered onto the back-side of the test section providing a0.5 cm2 each. Without any inlet restrictions in the micro-evaporator, significant parallel channel flowinstabilities, vapor back flow, and flow maldistribution led to high-amplitude and high-frequency tem-perature and pressure oscillations. Such undesired phenomena were then prevented by placing restric-tions at the inlet of each channel. High-speed flow visualization distinguished eight different operatingregimes of the two-phase flow depending on the tested operating conditions. Therefore, the preferredoperating regimes can be easily traced. In particular, flashing two-phase flow without back flow appearedto be the best operating regime without any flow and temperature instabilities.

� 2013 Elsevier Inc. All rights reserved.

1. Introduction and research motivation

The micro-electronics industry is searching for new solutions toextent the Complementary Metal–Oxide–Semiconductor (CMOS)performance trends over the next decades and to continue the pro-gress relative to Moore’s law (Moore, 1965). Three-dimensional(3D) integrated circuit stacked architecture with interlayer coolingseems to be an excellent opportunity to achieve this. In such sys-tem, the layers are placed one on top of each other and connectedusing vertical through-silicon vias (TSVs), which reduce the dis-tance for locally transporting data from one layer to another, mak-ing the process faster and generating less Joule heating. Due to highcosts and difficulties in designing and fabricating such 3D systems,the individual layers need to be tested beforehand in order todetermined their optimal geometries and operation. Therefore, asa part of the ‘CMOSAIC – 3D Stacked Architectures with InterlayerCooling’ project, the present experimental investigation focuses ona single layer of a future high-performance 3D stack of computerchip with 1 Tera nano-sized functional units compressed into1 cm3 (Sabry et al., 2011). Assuming that the individual layersare not interacting with each other, the results obtained for a singlelayer are being used as an input for designing a future 3D inte-grated circuit stacked architecture with interlayer cooling.

The 3D integration and continuously increasing heat fluxesdissipated by CPUs make the existing air-cooling technology

inadequate regarding its cooling capacity in respect to the giventemperature limit of 85 �C for CPUs, while dissipating an averageheat flux of about 100–150 W cm�2 and targeting the fluid temper-ature between 30 and 40 �C. Agostini et al. (2007) presented a com-prehensive state-of-the-art review of high heat flux coolingtechnologies including liquid jet impingement, single-phase liquidcooling, and two-phase flow boiling in copper or silicon micro-geometries. Recently, Marcinichen et al. (2010) and many otherresearchers proposed the multi-microchannel two-phase refriger-ant evaporation as an alternative to water cooling and a green solu-tion for cooling of computer blade servers and clusters, whichmight have up to 64 blades per rack cabinet, as well as their mem-ories. The application of two-phase cooling strategies introduces agreater degree of complexity in the modeling and design of suchsystems compared to single-phase techniques. However, these dif-ficulties are off set by the great increase in heat transfer coeffi-cients provided by the two-phase microchannel evaporators thatutilize the latent energy of the refrigerant to remove heat, provid-ing better axial temperature uniformity and reduced coolant flowrates, and thus much less pumping power, as stated by Marcini-chen et al. (2011).

However, a full understanding of two-phase flow boiling inmulti-microchannel cooling technologies, particularly flow insta-bilities, vapor back flow and flow maldistribution has yet to bemastered, thus creating a bottle neck in their commercial exploita-tion and indicating the need for continued research. Thus, visualobservation appears to be a powerful tool to study such phasechange phenomena and is the focus of this research. The presentpaper investigates the two-phase flow boiling of three dielectric

Nomenclature

Roman lettersa aspect ratio (–)B heater width (m)Bo boiling number (–)e expansion ratio (–)G mass flux (kg m�2 s�1)H depth (m)I current (A)L length (m)M mass flow rate (kg s�1)N number of channels (–)p pressure (Pa)q heat flux (W m�2)Re Reynolds number (–)STD standard deviation of temperature measurement (K)T temperature (�C)V voltage drop across the heater (V)W width (m)y direction perpendicular to the flow direction (m)z distance from the channel entrance (m)

Greek lettersDTin,sub inlet liquid subcooling (K)Dt time step (s)

e emissivity (–)u heat flux ratio (–)

Subscriptsamb ambientave averageb basech channelFS full scalef finh heaterIR infra-redin inletmax maximummin minimumout outletPx PyrexSi siliconsat saturationtape tapetotal total

S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189 177

fluids, namely R245fa, R236fa and R1234ze(E), in low aspect ratiochannels, each of them having a cross-sectional area of100 � 100 lm2 and following the global tendency towards minia-turization of electronics devices.

2. State-of-the-art review on the two-phase flow visualization

Over the past several years, visual observation of two-phaseflow patterns within single macro- and microchannels has been afocus of numerous experimental investigations (Ekberg et al.,1999; Yang and Shieh, 2001; Serizawa et al., 2002; Huo et al.,2004; Shiferaw et al., 2007; Shiferaw et al., 2009; Fu et al., 2008;Zhang and Fu, 2009; Saisorn et al., 2010; Martin-Callizo et al.,2010; Li and Peterson, 2005; Yen et al., 2006; Sobierska et al.,2006; Revellin et al., 2006; Ong and Thome, 2009; Ong and Thome,2011; Karayiannis et al., 2010; Tabatabai and Faghri, 2001; Galvisand Culham, 2012; Wang and Cheng, 2008; Celata et al., 2010).As a result, a variety of different types of flow pattern maps andtransition boundaries between particular flow patterns have beenproposed, for instance: in respect to superficial gas and liquidvelocities (Ekberg et al., 1999; Yang and Shieh, 2001; Serizawaet al., 2002; Huo et al., 2004; Shiferaw et al., 2007; Shiferawet al., 2009; Fu et al., 2008; Zhang and Fu, 2009; Saisorn et al.,2010; Martin-Callizo et al., 2010), or mass flux versus vapor quality(Revellin et al., 2006; Karayiannis et al., 2010; Ong and Thome,2009; Ong and Thome, 2011). Besides these flow pattern maps,some operational maps were developed based on the experimentalconditions, such as mass flux and heat flux. In this manner, Galvisand Culham (2012) have recently summarized their experimentalflow boiling patterns. Whereas, Wang and Cheng (2008) and Celataet al. (2010) categorized their experimental data into stable andunstable flows. Herein, a brief and concise survey on two-phaseflow visualization is provided, while a comprehensive state-of-the-art review on flow pattern maps is given in Thome (2010).

It has been shown that single channels have been the subject ofmost of these fundamental studies. Today, however, multi-micro-channels have gained particular attention in the micro-electronics

and power electronics industries, as explained earlier. Although,the number of parallel channels usually varies from 2 to over 60,two-phase flow visualization is commonly carried for only a fewchannels. Table 1 provides a summary of some previous experi-mental studies on the two-phase flow visualization in multi-micro-channel evaporators, among which only the visualizationperformed by Park and Thome (2009) was done for all the chan-nels. Therefore, the goal of this experimental investigation was tocarry out a 2D two-phase flow visualization of all the channelssimultaneously, as well as the inlet and the outlet headers.

3. Experimental set-up

3.1. Flow boiling test facility

The experimental flow boiling test facility contains of a closedloop of refrigerant, schematically presented in Fig. 1a. An oil-freemicropump coupled with a frequency controller is used to setthe desired liquid flow rate, which is measured by a Coriolis massflow meter with an accuracy of ±0.35% of the flow rate. A throttlevalve mounted before the test section is engaged to eliminateinstabilities related to the pumping system, keeping a trade-off be-tween an aperture of the valve and a rotational speed of the pump.In fact, the major pressure drop in the test loop is in the micro-evaporator, not in that valve. This control strategy (a supply pumpaccompanied by an inlet valve) was successfully utilized, for in-stance, by Zhang et al. (2011) in their tests in order to improvethe transient two-phase cooling performance. Downstream of theflow meter, to assure the desired condition for the test section in-let, there are a pre-heater/subcooler and an electrical pre-heater.The test section, described in detail in the following section, is a sil-icon micro-evaporator with a transparent Pyrex cover plate. Thepre-heater and the test section are heated by applying direct DCcurrent via two independent power sources. The voltage dropacross the pre-heater and the test section, and the current (in-di-rect measurement) are measured via the National Instrument dataacquisition system (NI DAQ). The micro-evaporator is inserted into

Table 1Some previous experimental studies on the two-phase flow visualization in multi-microchannel evaporators.

Reference Test fluid Testsection

Channel specifications Overallnumber ofchannels

Number ofchannelsvisualized

Main behaviors observed

Park and Thome(2009)

R134a,R236fa,andR245fa

Copper 30 mm-long, either467 � 4052 lm2 or199 � 756 lm2

(width � depth)

20 or 29 20 or 29 Two-phase flow stability was achieved by using the inlet orifices,which simultaneously imposed a different flow behavior

Hetsroni et al.(2001, 2002,2003, 2005)

Water Silicon 15 mm-long, 103 or129 lm of hydraulicdiameter

17, 21 or26

Below 20(commonly1)

Periodic annular flow and dry zone were observed, where thelatter led to periodic wetting and rewetting of the surface

Wu and Cheng(2003, 2004)

Water Silicon 30 mm-long, 82.8, 158.8,and 186 lm of hydraulicdiameter

8 or 15 1 Their experiments revealed temporal fluctuations of wall and fluidtemperatures, fluid pressures, and fluid mass flux

Wu et al. (2006) Water Silicon 60 mm-long, 72.7 lm ofhydraulic diameter

8 2 This paper is focused on the pressure drop and flow boilinginstabilities with decreasing mass fluxes at fixed power inputs

Wang et al.(2007, 2008)

Water Silicon 30 mm-long, 186 lm ofhydraulic diameter

8 From 1 to 5 Temperature and pressure oscillations for unstable two-phaseflow were studied. They showed that the pressure-drop-versus-flow-rate curve is predominant in the stability determination

Chen andGarimella(2006)

FC-77 Silicon 389 � 389 lm2 in size 24 1 The heat transfer coefficient drop and strong fluctuations of walltemperature at high heat fluxes due to the partial dryout werereported

Harirchian andGarimella(2009a,b,2010)

FC-77 Silicon From 96 to 707 lm ofhydraulic diameter

From 2 to63

1 The convective confinement number was proposed as a newtransition criterion for flow regime distinction in microchannelboiling

Bogojevic et al.(2009, 2011)

Water Silicon 15 mm-long, 273 lm-deep,150 lm-wide, spaced by100 lm fins

40 3 or 4,respectively

Two-phase flow instabilities were observed leading to remarkablepressure and temperature oscillations which depended on theheat and mass flux ratio and the inlet water temperature

178 S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189

a manifold (Fig. 1b), which is used to provide and remove fluidfrom the test section via the inlet and the outlet manifold plenums.In order to prevent leakage, 500 lm-diameter O-rings are placed inthe grooves around the manifold plenums. This manifold providesoptical access to both the top and bottom surfaces of the test sec-tion. The temperature of the refrigerant, as well as the pre-heaterwall temperature, are measured by Thermocoax 0.25 mm K-typethermocouples calibrated to within ±0.1 �C. The absolute pressuresat the inlet and the outlet manifold plenums are measured by10 bar absolute pressure sensors with a full-scale accuracy of±0.1%. Moreover, the pressure drop between the inlet to outletmanifold plenums is obtained by a differential pressure sensor

(a)Fig. 1. (a) Schematic of the flow boiling test facility, (b) bottom part of the test section mand (d) back-side of the PCB board with heaters and RTDs connectors.

with a ±0.023% full-scale accuracy. After the working fluid passesthrough the test section, it is directed to the tube-in-tube type heatexchanger where it is condensed and subcooled. The reservoir con-taining the working fluid is used to control the saturation pressurein the refrigerant circuit by adjusting its temperature using a RK 20KP LAUDA compact low-temperature thermostat. A 15 lm filter isinstalled on the facility to prevent any contaminants from enteringthe test section and blocking the microchannels.

The aim of the current research is to perform two-phase flowboiling and temperature visualizations over the entire micro-evap-orator area. In order to do that, the experimental test facility isequipped with an optical system comprising two high-speed

(b)

(c)

(d)

anifold, (c) front-side of the PCB board with HSEC8-130-01-S-DV-A edge connector,

S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189 179

cameras: a Photron Fastcam-Ultima APX camera placed above thetest section and a ThermaCAM SC3000 infra-red (IR) camera below.The Photron Fastcam-Ultima APX camera, having the highestspatial resolution of 1024 � 1024 pixels at 2000 fps, is coupledwith an AF Micro-Nikkor 105 mm 1:2.8D lens and an additionalring to increase magnification. An additional source of light is ap-plied to illuminate the test section. Its input is on the order ofabout 1% of the total heat flux applied (Szczukiewicz et al.,2012a, b, 2013), and thus it is assumed to be negligible. TheThermaCAM SC3000 high-speed IR camera, having a maximumframe rate of 900 fps (presently operated at 60 Hz), is equippedwith a close up lens LW 34/80, thus improving the measurementaccuracy over the small target areas by excluding unwantedbackground temperatures.

Use of IR thermography requires great attention to detail to getaccurate temperature measurements without reflection, emissiv-ity, and observation angle effects. Here, a novel in-situ pixel by pixeltechnique was developed to calibrate the raw IR image signals, andthus converting them into two-dimensional temperature fields ofthe test section surface in a manner that avoids most such prob-lems. In-situ IR camera calibration was performed with the flowloop by running single-phase liquid flow of a low-pressure refrig-erant, namely R245fa, and keeping adiabatic conditions in the testsection, so that any exchange of heat with the surrounding envi-ronment was allowed. Hence, a refrigerant temperature (here anaverage value of the temperature measured at the inlet and theoutlet manifold plenums, where the difference between themwas always within the measurement accuracy of the thermocou-ples employed, 60.1 K) could be assumed to be an IR temperature.The IR camera calibration was performed over the range of temper-atures from 10 and 60 �C with a temperature step of �5 �C. Whilecalibrating, the speed of the pump was set to nearly the maximumvalue to ensure uniform temperature distribution among thechannels. In the end, the IR temperatures that refer to junctiontemperatures commonly used in electronics devices are measuredwith an accuracy of ±0.2 �C, which provides tenfold accuracyimprovement compared to the manufacture’s value of ±2 �C. Moredetails on the present IR camera calibration, its accuracy, reliabil-ity, and reproducibility, can be found in Szczukiewicz et al.(2012a, b), Szczukiewicz (2012), Szczukiewicz et al. (2013).

3.2. Micro-evaporators

The test section, previously described in Szczukiewicz et al.(2012a, b, 2013), had overall dimensions 25.4 � 28 mm2. A380 lm-thick double-side polished silicon wafer was anodicallybonded to a 525 lm-thick transparent Pyrex cover plate. The sche-matic block diagram of the test section assembly is depicted inFig. 2a. Tests revealed that the micro-evaporator was able to with-stand absolute pressures up to 10 bars, thus indicating the highquality of the bonds. The channels, manufactured by a Deep Reac-tive Ion Etching (DRIE) process, are shown in Fig. 2b. They are10 mm-long with a 100 � 100 lm2 cross-section. The low aspectratio of the channel:

a ¼ Hch

Wch¼ 1 ð1Þ

minimizes the effect of flow stratification, due to the fact that thesurface tension forces overcome the gravitational forces (Thome,2010). The surface roughness of the channel bottom wall was mea-sured along its centerline using the non-contact optical phase shift-ing and white light vertical scanning interferometry technique,which gave a root mean squared value of 90 nm. There were 67channels separated by 50 lm-wide fins, which provided the 1 cm2

evaporating area commonly found in electronics devices. The fluidentered and exited the test section headers via, respectively,

1 mm-wide inlet and 2 mm-wide outlet slits. Two independent ser-pentine microheaters, made from aluminium of 1.5 lm thickness,were sputtered onto the back-side of the micro-evaporator, asshown in Fig. 2c, to simulate the power dissipated by active compo-nents in a 3D CMOS chip. Also, 50 lm-wide 4-wire resistance tem-perature detectors (RTDs) were plated onto the back-side inbetween the heater coils. these were calibrated in the same wayas the IR camera. DC current was applied to the micro-evaporatorvia a HSEC8-130-01-S-DV-A edge connector (SAMTEC) soldered tothe PCB board. Fig. 1c and 1d show the front- and back-side ofthe PCB. Then, this PCB was mounted on the side of the tested mi-cro-evaporator. As suggested by Hetsroni et al. (2001), Xu et al.(2005), a high-emissivity black matt tape (etape close to 1.0 withinthe present temperature range) was placed onto the back-side ofthe micro-evaporator to improve the accuracy of the infra-red tem-perature measurements and to minimize the IR transmissivity ofthe test section. The temperature distribution was uniform withthe spatial derivative of the temperature less than ±0.09 �C/pixel.

Snapshots from the high-speed visualization of two-phase flowof R245fa in two micro-evaporators, with and without inlet restric-tions, for the same experimental conditions are compared in Fig. 3.The flow direction is from left to right in all of presented images forboth flow and temperature. As expected, observation of the two-phase flow with evaporation of R245fa in the micro-evaporatorwithout any inlet restrictions revealed the presence of significantflow instabilities, vapor back flow and non-uniform flow distribu-tion among the channels. In order to assess the flow maldistribu-tion, the reader is referred to the electronic appendix of thispaper. Wu and Cheng (2003) pointed out that those phenomenalead to high-amplitude and high-frequency temperature and pres-sure oscillations. Such undesired phenomena might be preventedby using an inlet valve (Revellin et al., 2006), an inlet slit (Agostiniet al., 2008), or multiple restrictions at the entrance of each chan-nel (Kosar et al., 2006; Park, 2008; Park and Thome, 2009), whichrestrict and stabilize the two-phase flow. Therefore, the rectangu-lar restrictions of different expansion ratios (1.33, 2, and 4) wereplaced at the inlet of each channel. The expansion ratio down-stream of the inlet restriction is defined as follows:

ein;rest ¼Wch

Win;restð2Þ

Unfortunately, due to the unstable flow, the test section with-out any inlet restrictions failed before the experiment test matrixwas successfully accomplished. As explained by Park and Thome(2009), due to the flashing effect in the test section with the mi-cro-orifices, the boiling starts at a lower heat flux, and thus thewall-temperature distribution becomes more uniform and thewall-temperature overshoot for the onset of boiling is circum-vented by the vapor bubbles produced by the flashing. Therefore,inlet micro-orifices are certainly required in order to prolong thelifetime and extend the stable operation of the future 3D interlayercooling systems.

4. Experimental campaign

The geometrical specifications, experimental parameters anduncertainties are presented in Table 2. The uncertainties of themeasured values were obtained through the equipment calibra-tions, while the uncertainties of the derived parameters were cal-culated using the method developed by Kline and McClintock(1953). For all the tests, the outlet saturation temperature andthe inlet liquid subcooling were 31.5 ± 1 �C and 5.7 ± 1.5 K, respec-tively. The properties of the test fluids were obtained with the REF-PROP 8.0 software, the NIST Standard Reference Database 23. Fig. 4presents the summary of the experimental campaign that included

(a)

(b) (c)Fig. 2. (a) Schematic of the test section assembly (Sabry et al., 2011), (b) 100 � 100 lm2 channels with inlet restrictions of expansion ratio of ein,rest = 2, and (c) back-side ofthe silicon evaporator with the serpentine microheaters and RTDs.

180 S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189

4 different test sections (ein,rest = 1, 1.33, 2, and 4) and 3 refriger-ants: 2 low-pressure refrigerants R245fa, R236fa, and 1 medium-pressure new refrigerant R1234ze(E). The following comparisonsare then possible:

� – For R245fa flowing within the 2 test sections(ein,rest = 1 and 2)� – For R236fa flowing within the 3 test sections

(ein,rest = 1.33, 2, and 4)� – For ein,rest = 2 and the 3 fluids: R245fa, R236fa, and

R1234ze(E).

4.1. Experimental procedure

Firstly, the test facility was evacuated by a vacuum pump overat least 12 hours before filling it with the test fluid. The saturationpressure in the refrigerant circuit was set by adjusting the temper-ature in the LAUDA thermostat connected to the refrigerant tank.Then, the pressure and temperature were monitored to check fora presence of any contaminants in the loop. Once established, themass velocity was set through the frequency controller acting onthe micropump. Afterwards, the desired subcooling in the test sec-tion was set by means of the pre-heater/subcooler; however, theelectrical pre-heater could be also used to do this. Heat flux wasthen applied to the heater to initiate the transition to two-phaseflow in the test section. The experiments were performed at a con-stant heat flux being increased with a step of about 4 W cm�2,while the mass flux was gradually increased in steps of about200 kg m�2 s�1. The frequency response of the infra-red camerawas lower that the frequency of the bubble growth in the micro-channels, and thus the measurements were limited to steady state

conditions. The temperature, pressure, and voltage data were ac-quired by using the NI DAQ with the sampling rate of 1 kHz over1 min and then averaged. Due to the experimental protocol, thetime of 30 min between two test conditions was applied in allthe tests. As reported in Szczukiewicz et al. (2012a, b, 2013), a goodmeasurement reproducibility of the two-phase flow experimentswas obtained.

4.2. Single-phase flow tests

Prior to the boiling experiments, it is important to perform pre-validation tests to prove that the set-up is able to accurately mea-sure some single-phase liquid heat transfer coefficients. Such testswere performed for subcooled flow of 3 fluids within the 4 test sec-tions. Several different longitudinal distances, such as 0.5, 2.0 and5.0 mm, were analyzed (refer to Szczukiewicz, 2012; Szczukiewiczet al. (2012a, b, 2013)). The experimental Nusselt numbers werecompared with those predicted by Shah and London (1978) for-mula for laminar developing flow with a uniform heat flux bound-ary condition. Some discrepancies are revealed for high Reynoldnumbers, the reason being is that the prediction method doesnot consider downstream effects of the micro-orifices at the en-trance, which enhances the heat transfer with increasing the massvelocity, and thus the experimental Nusselt numbers are higherthan the predicted ones. Moreover, 3D conduction effects are nottaken into account in the data reduction, which may affect heattransfer results, as shown by Costa-Patry et al. (2012). The discrep-ancies seen for low Reynolds numbers, at 400, are due to the insuf-ficient experimental accuracy for such test conditions and the factthat the experimental facility was built to carry out two-phase flowexperiments.

(a)

(b)

Fig. 3. Visual observation of the two-phase flow boiling of R245fa forGch = 2035 kg m�2 s�1, qb = 36.5 W cm�2 in the micro-evaporator: (a) without anyinlet restrictions, and (b) with the 50 lm-wide, 100 lm-deep, and 100 lm-longmicro-orifices (ein,rest = 2) (Szczukiewicz et al., 2012b, 2013).

Table 2Operating conditions and experimental uncertainties.

Parameter Value Uncertainty

N (–) 67 –Lh (lm) 9765 ±5B (lm) 10,000 ±5Hch (lm) 100 ±5Wch (lm) 100 ±5Wf (lm) 50 ±5dSi (lm) 380 ±10dPx (lm) 525 ±10dh (lm) 1.5 ±5%Win,rest (lm) 25, 50, 75, and 100 (without any orifices) ±5Hin,rest (lm) 100 ±5Lin,rest (lm) 100 ±5Gch (kg m�2 s�1) 283–2370 8–30%a

I (A) 3.99–23.4 ±0.65%V (V) 0.52–2.51 ±0.53%TIR (�C) 29.9–57.1 ±0.2Tout,sat (�C) 31.5 ±1DTin,sub (K) 5.7 ±1.5pout (kPa) 188–597 ±0.1% FSDptotal (kPa) 2–122 ±0.023% FS

a Including unsteady flows.

Fig. 4. Summary of the experimental campaign, where � denotes the test case. (Forinterpretation of the references to color in this figure, the reader is referred to theweb version of this paper.)

S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189 181

At first, the heat emission by radiation was estimated based onthe Stefan–Boltzmann’s law for a black body (since etape is close to1.0 in the present temperature range) exposed to the surroundingwith a known temperature, Tamb, resulting in maximum 0.023 W ofheat emission by radiation. Therefore, the heat losses due to

radiation were shown to be negligible. Furthermore, in order toestimate the convective heat loss to the surrounding environment,single-phase energy balance tests for different mass fluxes, heatfluxes and inlet fluid subcoolings were performed. The fluid tem-perature within the loop was kept higher than the tested satura-tion temperature to maintain single-phase flow even at the hightemperatures. The measurements were done once the operationalsteady state condition was reached. As expected, the single-phaseliquid flow tests showed that the heat loss decreases rapidly withincreasing heat input, mass flow rate and inlet fluid subcooling andit becomes negligible (�3% or less) for the range of experimentalconditions corresponding to the two-phase flow experiments atthe tested subcooling (Szczukiewicz, 2012). When significant(low heat fluxes and mass fluxes), though, the heat loss to the envi-ronment is accounted for by correlating it as a function of the boil-ing and the Reynolds number, the average temperature of the testsection’s base, and the ambient temperature (respectively, Bo, Re,TIR,ave, and Tamb).

5. High-speed two-phase flow and temperature visualizations

Visual investigation into the two-phase flow dynamics in themicro-evaporator was carried out with the spatial resolution of1024 � 768 pixels at 2000 fps over a 4 s period with a field of viewthat includes the channels as well as the inlet and the outlet testsection’s plenums and slits. The single-phase liquid in the channelnear the entrance and the annular flow downstream in the channelare represented by the dark areas, where the two-phase flow mix-ture (bubbly and slug flows) appears brighter with a very shiny li-quid–vapor interface, as illustrated in Fig. 3. As mentioned above,the fluid flows from left to right side in all the presented images.

The temperature visualizations were conducted by means of thehigh-speed infra-red camera with the spatial resolution of320 � 240 pixels at 60 fps over a time of 1 min. Due to technical is-sues, an array of 100 � 100 IR temperature measurements coveredthe 1 cm � 1 cm heated area of the micro-evaporator. Thus, 600000 individual temperatures per second are acquired. The temper-ature data are obtained through the in-situ pixel by pixel calibrationof the infra-red camera described earlier that provides a tenfoldaccuracy improvement compared to the manufacture’s value inthe temperature range of the two-phase flow experiments.

182 S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189

5.1. Two-phase flow operational maps

In the current research, several new two-phase flow operationalmaps (Figs. 5–7) for 3 refrigerants, namely R245fa, R236fa, andR1234ze(E), flowing within the 4 multi-microchannel evaporators(ein,rest = 1, 1.33, 2, and 4) were developed for Tout,sat = 31.5 ± 1 �Cand D Tin,sub = 5.7 ± 1.5 K. Herein, it is worthwhile mentioning thatthe degree of inlet subcooling ensured no saturated flow at theinlet manifold’s plenum, while the refrigerant saturation

(a)

0 10 20 30 40 500

500

1000

1500

2000

2500

qb [W cm−2]

Gch

[kg

m−2

s−1

]

(a)

Fig. 5. Two-phase flow operational maps for the two-phase flow of R245fa in the microwide, 100 lm-deep, and 100 lm-long inlet restrictions (ein,rest = 2), where: – single-pmanifold’s outlet plenum, – single-phase flow followed by two-phase flow with backflow, – single-phase flow followed by two-phase flow without back flow, – two-phasection, – flashing two-phase flow with back flow, and – flashing two-phase flowcorrespond to the flow patterns shown earlier in Fig. 3. The black circles indicate the reprFig. 11 (temperature maps). (For interpretation of the references to color in this figure l

(a)

0 10 20 30 40 500

500

1000

1500

2000

2500

qb [W cm−2]

Gch

[kg

m−2

s−1

]

Fig. 6. Two-phase flow operational maps for the two-phase flow of R236fa in the mrestrictions (ein,rest = 1.33), and (b) the 25 lm-wide, 100 lm-deep, and 100 lm-long inlettest section with the vapor bubbles at the manifold’s outlet plenum operating regime,phase flow with back flow developing into jet flow, – jet flow, – single-phase flowtriggered by bubbles formed in the flow loop before the test section, – flashing two-pmost desirable). (For interpretation of the references to color in this figure legend, the r

temperature was chosen in order to provide a range of pressuressuitable for a single layer of a future high-performance 3D inte-grated circuit stacked architecture. These experimental results willbe then used as an input for a practical 3D computer chip, andtherefore the unified operational maps (independent of the exper-imental conditions, such as Tout,sat, DTin,sub, etc.) were not addressedin this work. The points represent the tested operating conditions,such as the channel mass flux, Gch, and the base heat flux, qb. Thefollowing definitions were employed here:

(b)

0 10 20 30 40 500

500

1000

1500

2000

2500

qb [W cm−2]

Gch

[kg

m−2

s−1

]

(h)

(a)

(c)

(b)

(e)(d)

(b)

(g)

-evaporator: (a) without any inlet restrictions (ein,rest = 1), and (b) with the 50 lm-hase flow – single-phase flow in the test section with the vapor bubbles at theflow, – unstable two-phase flow with back flow developing into jet flow, – jetse flow with back flow triggered by bubbles formed in the flow loop before the testwithout back flow (the most desirable). The data points marked by the red circlesesentative images of the operating regimes shown later in Fig. 8 (flow patterns) andegend, the reader is referred to the web version of this article.)

(b)

0 10 20 30 40 500

500

1000

1500

2000

2500

qb [W cm−2]

Gch

[kg

m−2

s−1

]

icro-evaporator with: (a) the 75 lm-wide, 100 lm-deep, and 100 lm-long inletrestrictions (ein,rest = 4), where: – single-phase flow, – single-phase flow in the– single-phase flow followed by two-phase flow with back flow, –unstable two-

followed by two-phase flow without back flow, – two-phase flow with back flowhase flow with back flow, and – flashing two-phase flow without back flow (theeader is referred to the web version of this article.)

S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189 183

Gch ¼M

NWchHchð3Þ

qb ¼ uIVLhB

ð4Þ

where u takes into account the voltage drop down the wiring be-tween the heater and the power supply.

Recently, Szczukiewicz et al. (2012a, b, 2013) reported on two-phase flow operating regimes of R245fa, R236fa and R1234ze(E) inthe micro-evaporator with the 50 lm-wide, 100 lm-deep, and100 lm-long micro-orifices (ein,rest = 2). The authors distinguished8 different operating regimes: (i) single-phase flow in the test sec-tion with the vapor bubbles at the manifold’s outlet plenum, (ii)single-phase flow followed by two-phase flow with back flow,(iii) unstable two-phase flow with back flow developing into jetflow, (iv) jet flow (two-phase flow is initiated only in few channels;vapor phase is push out of the channel creating a jet), (v)single-phase flow followed by two-phase flow without back flow(desirable operating regime), (vi) two-phase flow with back flowtriggered by bubbles formed in the flow loop before the testsection, (vii) flashing two-phase flow with back flow, and (viii)flashing two-phase flow without back flow (the most desirableoperating regime).

Their data were used to developed the two-phase flow opera-tional maps being shown in Fig. 5b and 7, where the following col-ors represent:

– single-phase flow,– single-phase flow in the test section with the vapor bubbles

at the manifold’s outlet plenum (i),– single-phase flow followed by two-phase flow with back

flow (ii),– unstable two-phase flow with back flow developing into jet

flow (iii),– jet flow (iv),– jet flow (iv),– two-phase flow with back flow triggered by bubbles formed

in the flow loop before the test section (vi),

(a)

0 10 20 30 40 500

500

1000

1500

2000

2500

qb [W cm−2]

Gch

[kg

m−2

s−1

]

Fig. 7. Two-phase flow operational maps for the two-phase flow of (a) R236fa, and (b) R1long inlet restrictions (ein,rest = 2), where: – single-phase flow, – single-phase flowfollowed by two-phase flow without back flow, – two-phase flow with back flow trigphase flow without back flow (the most desirable). The black circle indicates the representhe flow loop before the test section operating regime shown later in Fig. 8 (flow patternfigure legend, the reader is referred to the web version of this paper.)

– flashing two-phase flow with back flow (vii), and– flashing two-phase flow without back flow (viii).

The black circles in Fig. 5b and Fig. 7b indicate the representa-tive images of the above-mentioned operating regimes being illus-trated in Fig. 8 (flow patterns) and Fig. 11 (temperature maps) andthen described in the following subsection.

Moreover, the results for the other micro-evaporators(ein,rest = 1, 1.33, and 4) are added here in order to provide a com-prehensive view on the two-phase multi-microchannel phenom-ena (Fig. 5a and Fig. 6). The exact transitions between thesestates are not known, and thus their are not indicated on the maps.The notations from (i) to (viii) will be used later to refer to partic-ular operating regimes. As mentioned above, the micro-evaporatorwithout any inlet restrictions failed before its tests were success-fully accomplished. Therefore, in Fig. 5a, there are only eight pointson the map. This two-phase operational map is based on the se-lected data points. In some conditions in Fig. 6b, vapor bubblescould be flashed at the inlet of few channels, however, these exper-imental points could not be categorized as the flashing two-phaseflow with/without back flow, since the temperature trends are dif-ferent (for more details, refer to Szczukiewicz (2012)).

The following observations have been made:

� For R245fa flowing within the 2 test sections (ein,rest = 1 and 2)Fig. 5 shows the two-phase flow operational maps for R245faflowing within the 2 test sections: one without any inlet restric-tions and the second one with the inlet micro-orifices of theexpansion ratio of ein,rest = 2. As can be seen, the boiling startsat lower heat flux when using orifices. Furthermore, the operat-ing regime (viii) was not observed in the test section withoutany inlet restriction.� For R236fa flowing within the 3 test sections (ein,rest = 1.33, 2,

and 4)As shown in Fig. 6a in the range of the tested experimental con-ditions, neither (vii) nor (viii) operating regimes were observedfor R236fa flowing within the test section with the inletrestrictions of the expansion ratio of ein,rest = 1.33. These inlet

(b)

0 10 20 30 40 500

500

1000

1500

2000

2500

qb [W cm−2]

Gch

[kg

m−2

s−1

]

(f)

234ze(E) in the micro-evaporator with the 50 lm-wide, 100 lm-deep, and 100 lm-followed by two-phase flow with back flow, – jet flow, – single-phase flow

gered by bubbles formed in the flow loop before the test section, – flashing two-tatives images of the two-phase flow with back flow triggered by bubbles formed in

) and Fig. 11 (temperature map). (For interpretation of the references to color in this

(a) (b)

(c) (d)

(e) (f)

(g) (h)

Fig. 8. Two-phase flow operating regimes observed in the test section with the 50 lm-wide, 100 lm-deep, and 100 lm-long inlet restrictions (ein,rest = 2): (a) single-phaseflow in the test section with the vapor bubbles at the manifold’s outlet plenum, R245fa, (b) single-phase flow followed by two-phase flow with back flow, R245fa, (c) unstabletwo-phase flow with back flow developing into jet flow, R245fa, (d) jet flow, R245fa, (e) single-phase flow followed by two-phase flow without back flow, R245fa, (f) two-phase flow with back flow triggered by bubbles formed in the flow loop before the test section, R1234ze(E), (g) flashing two-phase flow with back flow, R245fa, and (h)flashing two-phase flow without back flow, R245fa.

184 S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189

restrictions do not provide enough pressure drop to generatesome flashed bubbles coming into the channels. For ein,rest = 2and 4, the zones where the operating regime (viii) was observedcorrespond to each other relatively well. For ein,rest = 2, thisoperating regime is observed above the channel mass flux than�1500 kg m�2 s�1 and base heat fluxes higher than�24 W cm�2 (Fig. 7a), whereas for ein,rest = 4, the transition linemoves to lower values of the mass flux (Fig. 6b).� For ein,rest = 2 and the 3 fluids: R245fa, R236fa, and R1234ze(E)

For R236fa, the operating regime (viii) was observed for massfluxes higher than �1500 kg m�2 s�1 and base heat fluxeshigher than �24 W cm�2, as depicted in Fig. 7a. For R245fa,there were only six data points recorded that showed this oper-ating regime (Fig. 5b), whereas for R1234ze(E), the two-phaseflow was stable for the mass fluxes higher than�900 kg m�2 s�1

and the base heat fluxes higher than �32 W cm�2 (Fig. 7b). It is

evident that the transition lines between particular operatingregimes are fluid-dependent. The flow of R236fa was stable overa wider range of parameters and the transitions in between theoperating regimes were more predictable, when comparing toR245fa and R1234ze(E).

5.2. Two-phase flow operating regimes

The flow remains in the single-phase regime in the micro-evap-orator for the lowest base heat flux. In some conditions, bubblesare observed at the manifold’s outlet plenum, where thesingle-phase flow exists in the microchannels, as seen in Fig. 8a.The transition to two-phase flow in the test section is a verydynamic process, and is highly unstable right after this occurs.The vapor bubbles come back to the inlet manifold’s and test sec-tion’s plenums. Therefore, the temporal and spatial non-uniformity

Fig. 10. Standard deviation of the image sequences over time provided by the two-phase flow of R245fa flowing in the test section with the inlet restrictions of theexpansion ratio of ein,rest = 2 for: (a) single-phase flow followed by two-phase flowwithout back flow operating regime, Gch = 2025 kg m�2 s�1, qb = 21.4 W cm�2 (desir-able operating regime), and (b) flashing two-phase flow without back flow operatingregime, Gch = 1895 kg m�2 s�1, qb = 47 W cm�2 (the most desirable operating regime).

S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189 185

of the flow inside the channels are observed (Fig. 8b and 8c). More-over, in Fig. 8b, 8c and 8g, the vapor back flow can be seen. For thejet flow operating regime (iii), which is an alternative repeatable intime, the flow transition from two-phase flow to jet flow is illus-trated in Fig. 8d. In case of the (iv) operating regime, the two-phaseflow is initiated only in a few channels and the vapor phase ispushed out of the channels creating liquid jets. The amount of va-por coming back to the inlet manifold’s plenum decreases whenincreasing the mass flow rate while the heat flux is kept constant.The vapor back flow is suppressed by the inlet restrictions. How-ever, the transition to two-phase flow does not happen at the samelongitudinal positions in all the channels, as shown in Fig. 8e. Theflow in the upper channels is still single-phase, while two-phaseflow has already appeared in the rest of the channels. The reasonfor this is the heat spreading into the copper manifold housingthe test section, which consequently reduces the heat flux to theflow.

As the heat flux is further increased, the operating regimes (vii)and (viii) are obtained and they are shown in Fig. 8g and 8h. Theflashing two-phase flow without back flow operating regime (viii)is the most desired operating regime for two-phase flow boiling inmulti-microchannels. The two-phase flow is initiated from thebeginning of the channel directly from flashed vapor bubbles.The bubble flow develops along the channel into slug and annularflows. Due to the flashing effect, the boiling starts at a lower valueof heat flux, thus, the wall-temperature distribution becomes moreuniform and the wall-temperature overshoot for the onset of boil-ing is circumvented (Park and Thome, 2009).

A time sequence of images in Fig. 9 illustrates some flashedbubbles are generated by the inlet restriction of the expansionratio of ein,rest = 4 (blue circle). They grow and become elongatedfurther downstream (Fig. 9b and 9c). In Fig. 9d to 9f, the situationrepeats (red circle). Furthermore, some isolated bubbles formed in

(a) (b)

(c) (d)

(e) (f)

(g) (h)

(i) (j)

Fig. 9. Flashed vapor bubbles of R236fa (white at left). Zoom-in into the inlet restriction of the expansion ratio of ein,rest = 4, Gch = 1311 kg m�2 s�1, qb = 13.9 W cm�2,Dt = 0.001 s. (For interpretation of the references to color in this figure, the reader is referred to the web version of this paper.)

(a)

z [mm]

y [m

m]

2 4 6 8 10

2

4

6

8

10

34

36

38

TIR [oC](b)

z [mm]

y [m

m]

2 4 6 8 10

2

4

6

8

10

38

39

40

41

TIR [oC]

(c)

z [mm]

y [m

m]

2 4 6 8 10

2

4

6

8

10

36

38

40

TIR [oC](d)

z [mm]

y [m

m]

2 4 6 8 10

2

4

6

8

10

38

40

42

44

TIR [oC]

(e)

z [mm]

y [m

m]

2 4 6 8 10

2

4

6

8

10

38

39

40

41 (f)

z [mm]

y [m

m]

2 4 6 8 10

2

4

6

8

10

34

35

36

TIR [oC]

(g)

z [mm]

y [m

m]

2 4 6 8 10

2

4

6

8

10

44

46

48

50

52

54TIR [oC]

(h)

z [mm]

y [m

m]

2 4 6 8 10

2

4

6

8

10

48

50

52

54

56

TIR [oC]

TIR [oC]

Fig. 11. Time-averaged temperature flow pattern maps for ein,rest = 2: (a) single-phase flow in the test section with the vapor bubbles at the manifold’s outlet plenum, R245fa,(b) single-phase flow followed by two-phase flow with back flow, R245fa, (c) unstable two-phase flow with back flow developing into jet flow, R245fa, (d) jet flow, R245fa, (e)single-phase flow followed by two-phase flow without back flow, R245fa, (f) two-phase flow with back flow triggered by bubbles formed in the flow loop before the testsection, R1234ze(E), (g) flashing two-phase flow with back flow, R245fa, and (h) flashing two-phase flow without back flow, R245fa. These temperature maps correspond tothe flow patterns in Fig. 8.

Table 3Width-wise temperature standard deviations for the four longitudinal channel positions (z = 2.0, 4.0, 6.0, 8.0 mm) and the eight representative images of the two-phase flowoperating regimes from Fig. 11.

Operating regime STD (K)

z (mm) 2.0 4.0 6.0 8.0Single-phase flow in the test section with the vapor bubbles at the manifold’s outlet plenum (i) 0.4 0.5 0.5 0.6Single-phase flow followed by two-phase flow with back flow (ii) 0.2 0.1 0.2 0.2Unstable two-phase flow with back flow developing into jet flow (iii) 0.3 0.3 0.3 0.3Jet flow (iv) 0.5 0.6 0.6 0.6Single-phase flow followed by two-phase flow without back flow (v) 0.2 0.2 0.2 0.2Two-phase flow with back flow triggered by bubbles formed in the flow loop before the test section (vi) 0.2 0.2 0.2 0.2Flashing two-phase flow with back flow (vii) 0.2 0.2 0.3 0.3Flashing two-phase flow without back flow (viii) 0.3 0.2 0.2 0.2

186 S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189

the piping before the test section pass through (Fig. 9g and 9h)without disturbing the flow in the micro-evaporator. In Fig. 9j,

no vapor back flow was observed (yellow circle). Those bubblesare created through a combine effect of fluid pre-heating and geo-

Table 4Detailed summary of the experimental campaign including the minimum and maximum junction temperatures of the micro-evaporator’s base (TIR,min and TIR,max), the maximumdissipated base heat flux (qb), and the corresponding total pressure drops (Dptotal).

Fluid Parameter

TIR,min (�C) TIR,max (�C) qb (W cm�2) Dptotal (kPa)

Expansion ratio ein,rest 1 1.33 2 4 1 1.33 2 4 1 1.33 2 4 1 1.33 2 4

R245fa 41.5 31.2 57.1 57.1 47.6 47.0 63.0 106.0R236fa 30.0 30.4 30.4 50.4 50.3 51.8 47.9 48.6 47.9 77.0 113.0 122.0R1234ze(E) 29.4 43.1 48.1 80.0

S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189 187

metrical components in the refrigerant loop. However, for some ofthe experimental conditions, vapor back flow can be triggered bythose bubbles (see Fig. 9f at left). Although, this phenomenon didnot have any significant effect on the two-phase flow in the testsection and the vapor back flow observed in these few cases (oper-ating regime (vi)) was not comparable to that occurring in theoperating regimes (ii), (iii), and (vii).

Fig. 10 illustrates the visual interpretation of the standard devi-ations obtained from 2729 images of flow visualization for operat-ing regimes (v) and (viii). The shiny areas correspond to the two-phase flow transition. In Fig. 10a, the two-phase starts at differentlongitudinal distances among the channels, whereas, in the othercase, the flow distributes uniformly with some flashed vapor bub-bles at the entrance of each channel. The flow develops into annu-lar flow faster in the middle channels. The flashing two-phase flowwithout back flow (viii) operating regime appeared to provide verygood flow uniformity in time and space (refer also to Fig. 3b and tothe electronic appendix of the present paper). The flow was stable,and thus it is the most desirable operating condition.

5.3. Temperature visualization

Fig. 11 illustrates several temperature maps of the test section’sbase obtained for different two-phase flow operating regimes. Oneneeds to notice that for visual reason, the temperature scales arenot the same (these maps with the same temperature scales andmodified intervals are demonstrated in Szczukiewicz (2012)). Forthe operating regimes from (i) to (vi), the temperature increasesalong the channel and slightly decreases at the channel outlet.The temperature increase towards the outlet is characteristic forthe single-phase flow. Whereas, for the operating regimes (vii)and (viii), the temperature decreases from inlet to outlet due tothe drop in the saturation pressure of the fluid. Table 3 presentsthe width-wise temperature standard deviations (STD) for the 4longitudinal channel positions (z = 2.0, 4.0, 6.0, 8.0 mm) and the8 representative images of the two-phase flow operating regimesfrom Fig. 11. The operating regimes (i) and (iv) provide the highestwidth-wise temperature standard deviations even up to 0.6 K. Incase of the operating regimes (ii), (iii), (vi), and (vii), the valuesof STD are relatively good, however, the two-phase flow is non-uniform and unstable with the vapor bubbles coming back to theinlet test section’s and manifold’s plenums. As shown, the temper-ature uniformity is the best for the operating regimes (v) and (viii).In the end, the flashing two-phase flow without back flow operat-ing regime (viii)), depicted in Fig. 11h, provided the best flow andtemperature uniformity and is the most desirable operatingregime.

Table 4 shows the minimum and maximum junction tempera-tures of the micro-evaporator’s base, respectively TIR,min and TIR,max,the maximum dissipated base heat fluxes, qb, as well as the corre-sponding maximum total pressure drops, Dptotal, for all the testcases. The maximum junction temperature of 57.1 �C was obtainedwhile dissipating the base heat flux 47.6 W cm�2 with 63 kPa ofthe total pressure drop (for the micro-evaporator without any inlet

restriction) or 47.0 W cm�2 of the base heat flux with 106 kPa ofthe total pressure drop (for the test sections with the micro-ori-fices). These are not however the critical heat fluxes, which wereavoided here to not risk the burnout of the test sections.

6. Conclusions and final remarks

The following conclusions can be drawn from the currentresearch:

� The importance of the overall visual inspection in multi-micro-channel experiments to study phase change phenomenon hasbeen shown.� Three different micro-orifice expansion ratios (ein,rest = 1.33, 2,

and 4) have been tested and the obtained results were com-pared with the test section that did not have any inletrestrictions.� Micro-orifices having the expansion ratio of 2 and 4 successfully

suppressed flow instability, vapor back flow and significantlyimproved flow uniformity among the channels in the range ofthe tested operating conditions. Hence, the use of micro-orificesis essential for two-phase cooling in the micro-electronicsindustry in order to ensure the stable operation of future 3Dinterlayer cooling systems.� Six two-phase flow operational maps for multi-microchannel

evaporators have been developed identifying stable and unsta-ble operating regimes according to experimental conditions,such as the channel mass flux and the base heat flux.� The two-phase flow test experimental conditions were catego-

rized into eight operating regimes, among which the flashingtwo-phase flow without back flow (viii) operating regime wasfound to provide the best flow and temperature uniformity intime and space.� The transitions from one regime to another were fluid-depen-

dent. Among the test fluids, refrigerant R236fa is proposed hereto be used in the future 3D high-performance computer chips.The reason is that this fluid was stable over a wider range ofexperimental conditions and the transitions in between theoperating regimes were more predictable compared to R245faand R1234ze(E).� The maximum junction temperature of 57.1 �C was obtained

while dissipating the base heat flux 47.6 W cm�2 with63 kPa of the total pressure drop (for the micro-evaporatorwithout any inlet restriction) or 47.0 W cm�2 of the base heatflux with 106 kPa of the total pressure drop (for the test sec-tions with the micro-orifices). For all the test cases the tem-perature was below the given temperature limit of 85 �C forCPUs, while dissipating half of the heat flux being targeted.Thus, further research is needed to achieve the milestones ofthe CMOSAIC project. The fact that needs to be highlightedhere is the following: the junction temperature of the micro-evaporator can still increase more than 25 �C before the givenlimit is reached.

188 S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189

Acknowledgements

This research is funded by the Nano-Tera RTD project CMOSAIC(Ref. 123618) financed by the Swiss Confederation and scientifi-cally evaluated by SNSF. The authors thank the MicroelectronicSystems Laboratory (LSM, EPFL) for manufacturing the test sec-tions, Advanced Thermal Packaging group (IBM Zürich) for bond-ing, and the workshop of the Heat and Mass Transfer Laboratory(LTCM, EPFL) for the work done on the experimental facility. Wewish to further thank Honeywell Inc. (Samuel Yana Motta) for pro-viding the R1234ze(E) test fluid.

Appendix A

The electronic appendix of the current paper consists of thehigh-speed videos of the two-phase flow of R245fa in the 2 mul-ti-microchannel evaporators: one with the 50 lm-wide, 100 lm-deep, and 100 lm-long inlet restrictions (ein,rest = 2) and one with-out any inlet restrictions (ein,rest = 1). The videos are displayed inslow motion at a frame rate of 30 fps.

Appendix B. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.ijheatfluidflow.2013.03.006.

References

Agostini, B., Fabbri, M., Park, J., Wojtan, L., Thome, J.R., Michel, B., 2007. State of theart of high heat flux cooling technologies. Heat Transfer Eng. 28 (4), 258–281.

Agostini, B., Thome, J.R., Fabbri, M., Michel, B., Calmi, D., Kloter, U., 2008. High heatflux flow boiling in silicon multi-microchannels – Part I: Heat transfercharacteristics of refrigerant R236fa. Int. J. Heat Mass Transfer 51, 5400–5414.

Bogojevic, D., Sefiane, K., Walton, A.J., Lin, H., Cummins, G., 2009. Two-phase flowinstabilities in a silicon microchannels heat sink. Int. J. Heat Fluid Flow 30, 854–867.

Bogojevic, D., Sefiane, K., Walton, A.J., Lin, H., Cummins, G., Kenning, D.B.R.,Karayiannis, T.G., 2011. Experimental investigation of non-uniform heatingeffect on flow boiling instabilities in a microchannel-based heat sink. Int. J.Therm. Sci. 50, 309–324.

Celata, G.P., Saha, S.K., Zummo, G., Dossevi, D., 2010. Heat transfer characteristics offlow boiling in a single horizontal microchannel. Int. J. Therm. Sci. 49, 1086–1094.

Chen, T., Garimella, S.V., 2006. Measurement and high-speed visualization of flowboiling of a dielectric fluid in a silicon microchannel heat sink. Int. J. Heat MassTransfer 32, 957–971.

Costa-Patry, E., Nebuloni, S., Olivier, J., Thome, J.R., 2012. On-chip two-phase coolingwith refrigerant 85 lm-wide multi-microchannel evaporator under hot-spotconditions. IEEE Trans. Compon. Packag. Manufact. Technol. 2 (2), 311–320.

Ekberg, N.P., Ghiaasiaan, S.M., Abdel-Khalik, S.I., Yoda, M., Jeter, S.M., 1999. Gas-liquid two-phase flow in narrow horizontal annuli. Nucl. Eng. Des. 192, 59–80.

Fu, X., Qi, S.L., Zhang, P., Wang, R.Z., 2008. Visualization of flow boiling of liquidnitrogen in a vertical mini-tube. Int. J. Multiphase Flow 34, 333–351.

Galvis, E., Culham, R., 2012. Measurements and flow pattern visualizations of two-phase flow boiling in single channel evaporators. Int. J. Multiphase Flow 42, 52–61.

Harirchian, T., Garimella, S.V., 2009a. Effects of channel dimension, heat flux, andmass flux on flow boiling regimes in microchannels. Int. J. Multiphase Flow 35,349–362.

Harirchian, T., Garimella, S.V., 2009b. The critical role of channel cross-sectionalarea in microchannel. Int. J. Multiphase Flow 35, 904–913.

Harirchian, T., Garimella, S.V., 2010. A comprehensive flow regime map formicrochannel flow boiling with quantitative transition criteria. Int. J. HeatMass Transfer 53, 2694–2702.

Hetsroni, G., Mosyak, A., Segal, Z., 2001. Nonuniform temperature distribution inelectrical devices cooled by flow in parallel microchannels. IEEE Trans. Compon.Packag. Technol. 24, 16–23.

Hetsroni, G., Mosyak, A., Segal, Z., Ziskind, G., 2002. A uniform temperature heat sinkfor cooling of electroni devices. Int. J. Heat Mass Transfer 45, 3275–3286.

Hetsroni, G., Mosyak, A., Segal, Z., Pogrebnyak, E., 2003. Two-phase flow patterns inparallel micro-channels. Int. J. Multiphase Flow 29 (3), 341–360.

Hetsroni, G., Mosyak, A., Pogrebnyak, E., Segal, Z., 2005. Explosive boiling of water inparallel micro-channels. Int. J. Multiphase Flow 31 (4), 371–392.

Huo, X., Chen, L., Tian, Y.S., Karayiannis, T.G., 2004. Flow boiling and flow regimes insmall diameter tubes. Appl. Therm. Eng. 24, 1225–1239.

Karayiannis, T.G., Shiferaw, D., Kenning, D.B.R., Wadekar, V.V., 2010. Flow patternsand heat transfer for flow boiling in small to micro diameter tubes. HeatTransfer Eng. 31 (4), 257–275.

Kline, S.J., McClintock, F.A., 1953. Describing uncertainties in single-sampleexperiments. Mech. Eng. 75, 3–8.

Kosar, A., Kuo, Ch.-J., Peles, Y., 2006. Suppression of boiling flow oscillations inparallel microchannels by inlet restrictions. J. Heat Transfer 128, 251–260.

Li, J., Peterson, G.P., 2005. Boiling nucleation and two-phase flow patterns in forcedliquid flow in microchannels. Int. J. Heat Mass Transfer 48, 4797–4810.

Marcinichen, J.B., Thome, J.R., Michel, B., 2010. Cooling of microprocessors withmicro-evaporation: a novel two-phase cooling cycle. Int. J. Refrigeration 33,1264–1276.

Marcinichen, J., Olivier, J., Thome, J.R., 2011. Reasons to Use Two-Phase RefrigerantCooling. <www.electronics-cooling.com>.

Martin-Callizo, C., Palm, B., Owhaib, W., Ali, R., 2010. Flow boiling visualization of R-134a in a vertical channel of small diameter. J. Heat Transfer 132, 1–8.

Moore, G.E., 1965. Cramming more components onto integrated circuits. Electronics38, 114–117.

Ong, Ch. L., Thome, J.R., 2009. Flow boiling heat trasnfer of R134a, R236fa andR245fa in a horizontal 1.030 mm circular channel. Exp. Therm. Fluid Sci. 33,651–663.

Ong, Ch. L., Thome, J.R., 2011. Macro-to-microchannel transition in two-phase flow:Part 1 – Two-phase flow patterns and fil thickness measurements. Exp. Therm.Fluid Sci. 35, 37–47.

Park, J.E., 2008. Critical heat flux in multi-microchannel copper elements with lowpressure refrigerants. PhD thesis, Ecole Polytechnique Federale de Lausanne,<http://biblion.epfl.ch/EPFL/theses/2008/4223/EPFL_TH4223.pdf>.

Park, J.E., Thome, J.R., Michel, B., 2009. Effect of inlet orifices on saturated CHF andflow visualisation in multi-microchannel heat sinks. In: 25th IEEE SEMI-THERMSymposium.

Revellin, R., Dupont, V., Ursenbacher, T., Thome, J.R., Zun, I., 2006. Characterizationof two-phase flows in microchannels: optical measurement technique and flowparameter results for R-134a in a 0.5 mm channel. Int. J. Multiphase Flow 32 (7),755–774.

Sabry, M.M., Sridhar, A., Atienza, D., Temiz, Y., Leblebici, Y., Szczukiewicz, S.,Borhani, N., Thome, J.R., Brunschwiler, T., Michel, B., 2011. Towards thermally-aware design of 3D MPSoCs with inter-tier cooling. In: Design, Automation andTest in Europe.

Saisorn, S., Kaew-On, J., Wongwises, S., 2010. Flow pattern and heat transfercharacteristics of R-134a refrigerant during flow boiling in a horizontal circularmini-channel. Int. J. Heat Mass Transfer 53, 4023–4038.

Serizawa, A., Feng, Z., Kawara, Z., 2002. Two-phase flow in microchannels. Exp.Therm. Fluid Sci. 26, 703–714.

Shah, R.K., London, A.L., 1978. Laminar Flow Forced Convection in Ducts. AcademicPress.

Shiferaw, D., Huo, X., Karayiannis, T.G., Kenning, D.B.R., 2007. Examination of heattransfer correlations and a model for flow boiling of R134a in small diametertubes. Int. J. Heat Mass Transfer 50, 5177–5193.

Shiferaw, D., Karayiannis, T.G., Kenning, D.B.R., 2009. Flow boiling in a 1.1 mm tubewith R134a: experimental results and comparison with model. Int. J. Therm. Sci.48, 331–341.

Sobierska, E., Kulenovic, R., Mertz, R., Groll, M., 2006. Experimental results offlow boiling of water in a vertical microchannel. Exp. Therm. Fluid Sci. 31, 111–119.

Szczukiewicz, S., 2012. Thermal and visual operational characteristics of multi-microchannel evaporators using refrigerants. PhD thesis, Ecole PolytechniqueFederale de Lausanne, <http://infoscience.epfl.ch/record/182692/files/EPFL_TH5594.pdf>.

Szczukiewicz, S., Borhani, N., Thome, J.R., 2012a. Two-phase flow boiling in a singlelayer of future high-performance 3D stacked computer chips. In: 13th IEEEIntersociety Conference on Thermal and Thermomechanical Phenomena inElectronic Systems.

Szczukiewicz, S., Borhani, N., Thome, J.R., 2012b. Two-phase heat transfer and high-speed visualization of refrigerant flows in 100 � 100 lm2 silicon multi-microchannels. In: ECI 8th International Conference on Boiling andCondensation Heat Transfer.

Szczukiewicz, S., Borhani, N., Thome, J.R., 2013. Two-phase heat transfer and high-speed visualization of refrigerant flows in 100 � 100 lm2 silicon multi-microchannels. Int. J. Refrigeration 36, 402–413.

Tabatabai, A., Faghri, A., 2001. A new two-phase flow map and transition boundaryaccounting for surface tension effects in horizontal miniature and micro tubes.Trans. ASME 123, 958–968.

Thome, J.R., 2010. Chapter 20: Wolverine Engineering Data Book III.<www.wlv.com>.

Wang, G., Cheng, P., 2008. An experimental study of flow boiling instability in asingle microchannel. Int. Commun. Heat Mass Transfer 35 (10), 1229–1234.

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

Wang, G., Cheng, P., Bergles, A.E., 2008. Effects of inlet and outlet configurations onflow boiling instabilities in parallel microchannels. Int. J. Heat Mass Transfer 51,2261–2281.

Wu, H.Y., Cheng, P., 2003. Visualization and measurement of periodic boiling insilicon microchannels. Int. J. Heat Mass Transfer 46, 2603–2614.

Wu, H.Y., Cheng, P., 2004. Boiling instability in parallel silicon microchannels atdifferent heat flux. Int. J. Heat Mass Transfer 47, 3631–3641.

S. Szczukiewicz et al. / International Journal of Heat and Fluid Flow 42 (2013) 176–189 189

Wu, H.Y., Cheng, P., Wang, H., 2006. Pressure drop and flow boiling instabilities insilicon microchannel heat sinks. J. Micromech. Microeng. 16, 2138–2146.

Xu, J., Shen, S., Gan, Y., Li, Y., Zhang, W., Su, Q., 2005. Transient flow pattern basedmicroscale boiling heat transfer mechanisms. J. Micromech. Microeng. 15 (6),1344.

Yang, Ch.-Y., Shieh, Ch.-Ch., 2001. Flow pattern of air-water and two-phase R-134ain small circular tubes. Int. J. Multiphase Flow 27, 1163–1177.

Yen, T.-H., Shoji, M., Takemura, F., Suzuki, Y., Kasagi, N., 2006. Visualization ofconvective boiling heat transfer in single microchannels with different shapedcross-sections. Int. J. Heat Mass Transfer 49, 3884–3894.

Zhang, P., Fu, X., 2009. Two-phase flow characteristics of liquid nitrogen in verticallyupward 0.5 and 1.0 mm micro-tubes: visualization studies. Cryogenics 49, 565–575.

Zhang, T., Wen, J.T., Peles, Y., Catano, J., Zhou, R., Jensen, M.K., 2011. Two-phaserefrigerant flow instability analysis and active control in transient electronicscooling systems. Int. J. Multiphase Flow 37, 84–97.