Nanofluid spray cooling characterization combiningphase Doppler anemometry with high-speed

visualization and thermography

Miguel Silva Figueiredomiguel.s.figueiredo@tecnico.ulisboa.pt

Instituto Superior Técnico, Universidade de Lisboa, Portugal

October 2020

Abstract

The spray-wall interactions depict thermo fluid dynamics phenomena exploited in a vast range of ap-plications, from internal combustion engines, electronic components cooling, HVAC, to numerous medicaland industrial applications. Additionally, climate change awareness cries for the urgent development ofmore efficient and more climate-friendly cooling systems. For these reasons, the scientific community hasbeen showing an interest in uncovering nanofluid possible applications. As they possess superior thermalconductivity and overall better thermal cooling performance, their potential use as a coolant has becomea great promise.

In this order, nanofluids with gold and silver nanoparticles, with different geometries and concentra-tions ranging from 0.1 - 1 (wt.%) were atomized and characterized. The study performed on this disserta-tion adopted a novel combination of three techniques little exploited in the literature. The phase Doppleranemometry system and the synchronization of a high-speed camera with a high-speed thermographicinfrared camera were used to describe spray impact onto the heated surface.

According to the results, a decreased surface tension and higher impingement distance favour heattransfer from the wall, as the wetted area increased. Moreover, changes in the thermophysical propertieswere noted with the addition of nanoparticles, when compared to the base fluid. Although their presencedid not affect the spray dynamics, an increase of 9.8% to 21.9% of the maximum heat flux was notedduring impact when compared to the base fluid.

Keywords: Spray cooling, Nanofluid, Time Resolved Infrared Thermography, Phase DopplerAnemometry

1. IntroductionOver the last few years, the development of theelectronic systems, mainly in processing electron-ics, urges for the development of high heat flux re-moval systems. Cooling systems must assure thegood functioning and endurance of these systemsthat are becoming more power concentrated due toaggressive miniaturization. The use of liquids hasbecome an inevitable response, whether in sin-gle or two-phase liquid cooling, as heat dissipationfrom supercomputer chips approached 100W/cm2

[1, 2].The most used liquid cooling systems come from

pool boiling, channel or microchannel flow boiling,jet-impingement and spray cooling. Each approachhas its own advantages, nevertheless, spray cool-ing is considered by some authors as the most ad-vantageous (e.g. [1]). In ideal conditions, spray

cooling achieves a heat flux removal in the or-der of 1200W/cm2, an order of magnitude higherthan pool boiling, which only achieves 140W/cm2.When compared with jet-impingement, spray cool-ing offers better spatial cooling uniformity, that oth-erwise would be detrimental to some devices, re-sulting in higher cooling efficiencies and lower liq-uid consumption [2, 3].

In a simple way, spray cooling is obtained bythe impact of the spray on the surface/element tocool. The spray is globally composed by numerousdroplets within a wide range of sizes and velocities,which are generated at the nozzle, where the liquidbreaks up due to instabilities, caused by its mo-mentum. A spray, however, cannot be simply mod-elled as a sum of individual droplets, since thereare numerous interactions between the droplets,with the air, and at impact with droplets previously

1

spread on the surface as well as with a possible de-posited liquid film, also resulting from earlier sprayinjection. However, spray cooling, due to its com-plexity, still lacks theoretical modelling when com-pared to other cooling techniques. It is consideredthat spray cooling is limited by the cooling effec-tiveness of the conventional refrigerants, such aswater, engine oil or ethylene [4].

Therefore, large interest by the scientific com-munity regarding the scrutiny of nanofluid applica-tions in cooling over the last few years is easilynoticed. These nanofluids come with the promiseof achieving higher heat transfer efficiencies, dueto their enhanced thermal conductivity, when com-pared with conventional cooling fluids, whose prop-erties largely enhance their cooling performance[5, 6].

Nanofluids are composed of solid particles thatare dispersed in a base liquid (e.g. Water, Refrig-erant Oils). These particles can be metallic (e.g.Aluminum, Gold, Silver) or nonmetallic (e.g. Car-bon and Nitrides), whose dimensions are in thenanometer range (10−100nm) [7]. Since nanopar-ticles have higher thermal conductivities, at leasttwo orders of magnitude higher than their base liq-uid, they enhance the fluid’s overall thermal con-ductivity. Additionally, other fluid-dynamic char-acteristics are stated in various studies, such asBrownian movement and higher fluid stability thatshould prevent particle settlement [6].

Nevertheless, controversial conclusions regard-ing the increase of the thermal conductivity andits contribution to enhance or worsen heat trans-fer can be found in the literature. In one hand, thismeans that more studies and more detailed analy-sis are still needed [6, 4]. On the other hand, thereare other properties such as viscosity and localwettability, which can be strongly affected by theaddition of the nanoparticles to the base fluids, al-tering the hydrodynamic characteristics of the flow,in complex processes which are still far to be accu-rately described.

In this regard, in order to diminish the existinggaps in this topic, this study is aimed at correlat-ing the nanoparticle presence in the base fluid andits effect on thermophysical properties. Therefore,its influence on the atomization process and heattransfer is analyzed at two impact distances andat two initial surface temperatures. This study isa followup to a previous study [8], and it consid-ers nanofluids with different nanoparticle chemicalelement, geometry and concentration. The atom-ization and heat transfer process are character-ized using a novel combination of three techniques:phase Doppler anemometry and high speed visual-ization imaging, coming from a high speed cameraand a thermographic camera.

2. Background2.1. Nanofluid propertiesTo evaluate the thermal conductivity, the classi-cal computation approach will be adopted. Con-sidering well-dispersed solid spherical particles ona continuum medium, the thermal conductivity isgiven by:

Knf =2Kbf +Kp + 2φ (Kp −Kbf )

2Kbf +Kp − φ (Kp −Kbf )Kbf (1)

where φ is the volume fraction of the solid parti-cles, Kbf and Kp the base fluid and solid particlethermal conductivity, respectively, and consideringthem as bulk materials.

However, when considering different particleshapes, the introduction of an empirical shape fac-tor (n) that relates the particle sphericity is needed,resulting in Equation (2). The shape factor isthe ratio of the surface area of the non-sphericalnanoparticle to that of an equivalent sphericalnanoparticle with the same volume (n = S/S

′).

Knf =Kp + (n− 1)Kbf − (n− 1)φ (Kbf −Kp)

Kp + (n− 1)Kbf + φ (Kbf −Kp)Kbf

(2)The specific heat (cp) is another fundamental

property. It is mainly estimated based on heat equi-librium. According to Xuan and Roetzel (2000), cpcan be estimated as:

cpnf=

(1− φ) (ρcp)bf + φ (ρcp)p(1− φ) ρbf + φρp

(3)

Finally, the specific mass (ρ) is calculated as:

ρnf = (1− φ) ρbf + φρp (4)

3. Experimental setup and procedures3.1. Experimental SetupThe experimental arrangement used in the presentwork is schematically represented in Figure 1. Thefluids under study were atomized and character-ized by two distinct system arrangements. The un-derlined area of the scheme represents the maincomponents of the experimental arrangement thatincludes the atomizer and the impact surface.

The atomizer (fig. 1, index 1) is composed oftwo parts: the swirl nozzle and the nozzle supportthat receives the liquid tangentially. The swirl noz-zle has a discharge orifice of 0.42 mm in diameterand two opposing tangential ports with a squaredshaped cross-section of 0.6 × 0.6mm2. Detailedgeometrical specifications can be found in [8].

The continuum liquid stream comes from a 3dm3

cylindrical reservoir (fig. 1, Reservoir 1), that ispressurized by air. The pressurization (relative)

2

S

-

+IR

Cam

era

HS Camera

DAQRese

rvoi

r1

Reservoir2

PC 1PC 2

Ar-Ion LaserTransmitting Optics

Receiving Optics and

Detectors

Particle and Flow Processor

2

11

5

10

6

7

8

9

9

3

4

Setup 1 Setup 2

Figure 1: Experimental installation scheme: (1) Atomizer; (2)Air pressure regulator; (3) Manometer; (4) Temperature sensor;(5) Impact surface; (6) Power supply; (7) Electric cables; (8)Solenoid valve; (9) Manual valve; (10) Light source

was maintained at 87psi (±0.5psi) by a IR1M Mon-nier compressed air regulator from Spirax Sarco(fig. 1, index 2) and monitored by a manometer(fig. 1, index 3).

Additionally, a K-type thermocouple (C03-K fromOmega, fig. 1, index 4) was mounted on Reser-voir 1 and it was connected to a data acquisitionboard DT9828 from Data Translation (fig. 1, DAQ)to measure the liquid’s initial temperature.

A Phase Doppler Anemometry - PDA systemwas used to characterize the velocity and size ofthe droplets that constitute the spray, in variouspoints of interest. This system was utilized inSetup 1 and it is composed by a 300mW −400mWAr-Ion Laser, from Spectra-Physics, a transmit-ting/receiving optics connected to a particle andflow processor (BSA P80, from Dantec Dynamics).

In Setup 2, the heat exchange between the sprayand the heated surface was analyzed. A metal-lic base (fig. 1, index 5) paired with a thin stain-less steel (AISI 304) surface was used to char-acterize the spray impact phenomena. This sur-face had a thickness of 20µm and was heatedby Joule effect using a continuous current powersupply (HP 6274B DC, fig. 1, index 6). Addition-ally, as this surface was very thin, the tempera-ture variations at its inferior face were captured bya high-speed infrared (IR) thermographic camera(Onca-MwIR-InSb-320, from Xenics). Moreover, asolenoid valve (SV3108 from Omega, fig. 1, index8) was mounted before the nozzle to improve therepeatability of the tests performed with the ther-mographic camera.

In both setups, a high-speed (HS) camera(Phantom v4.2) was mounted perpendicularly tothe spray, allowing not only the capture of the freespray (used in Setup 1) but also the spray impactonto the heated surface (used in Setup 2). Alongwith this camera, a light source of 50 Watts with adiffusing glass (fig. 1, index 10) was mounted on

the opposite side of the spray to improve the cap-ture contrast. It is important to emphasize that both(HS and IR) cameras were synchronized for sprayimpact on Setup 2.

To finalize, after atomization the working fluidswere redirected to a reservoir (fig. 1, Reservoir 2),where they are filtered and reutilized afterwards. Itis worth mentioning that the most relevant thermo-physical properties of all the nanofluids used herewere evaluated before and after the experimentaltests (i.e. before and after atomization). Changesin these properties were observed to be negligible,which supports the argument that there were nosignificant particle losses in the liquid feeding sys-tem or the atomizer, in agreement with a previousstudy [8].

3.2. Nanofluid Preparation and CharacterizationOne of the key aspects of the preparation of ananofluid is to guarantee its homogeneity andstability. In this regard, the addition of a sur-factant is advised and, in this context, the useof Cetyltrimethylammonium bromide (CTAB) wasadopted. However, one of the downsides of theuse of surfactants, including CTAB, is their impacton the liquid’s surface tension. Additionally, as dif-ferent nanoparticle concentrations were going tobe used in this study, different surfactant concen-trations were required to assure the homogeneityof the nanofluids at higher nanoparticles concen-trations. Therefore, the first step was to establisha range of surfactant concentrations and analyzetheir effect on Deionized (DI) Water surface ten-sion. To that end, different concentrations of DIWater-CTAB mixtures ranging from 0.01 − 1(wt.%)were prepared and sonicated. Their surface ten-sion was measured, and their values are repre-sented in Figure 2.

0 , 0 0 , 2 0 , 4 0 , 6 0 , 8 1 , 03 0

3 5

4 0

4 5

5 0

5 5

Surfa

ce Te

nsion

(mN/m

)

S u r f a c t a n t C o n c e n t r a t i o n ( w t . % )

S e l e c t e d C o n c e n t r a t i o n R a n g e

Figure 2: Surface tension variation with CTAB concentration inDI Water

From Figure 2, a rapid decrease of the surfacetension is evident for very low surfactant concen-trations (e.g. < 0.04(wt.%)). However, the surfacetension value stays constant above 0.04(wt.%),with variations lower than 3.2%. Therefore, thesurfactant concentration range was fixed between

3

0.05 - 1(wt.%).Afterwards, the nanofluids were prepared at

the Structural Chemistry Center of the Depart-ment of Chemical Engineering at IST. These weresupplied with different nanoparticle concentrationsin the range of 0.1 - 1 (wt.%), mixed with DIWater containing surfactant. Thereafter, thesewere sonicated using a tip sonicator processor(UP200Ht from Hielscher), for 20 minutes on av-erage at a mean amplitude of 40%. This ensureda good nanofluid homogeneity, as the majority ofthe nanofluids under study did not show apparentnanoparticle deposition for over a week. However,to avoid nozzle clogging, they were sonicated ex-actly before use.

The main characteristics and thermophysicalproperties for all the fluids used in this study, mea-sured as described in the following paragraphs, aresummarized in Table 1.

The thermal conductivity, specific mass and spe-cific heat, were theoretically calculated based ontheir nanoparticle chemical element and concen-tration, following the Equations (1) to (4).

Viscosity was measured by the Structural Chem-istry Center of the Department of Chemical Engi-neering, using a rheometer (A instruments ARI 500ex) at room temperature (20◦C).

Surface tension was measured with an opticaltensiometer THETA, from Attension. The pendantdrop method was used under controlled room tem-peratures (20 ± 3◦C). For each solution, 15 mea-surements were performed and averaged.

The refractive index was also evaluated since itis a fundamental input parameter for the PDA sys-tem. This property was evaluated with an Abbe re-fractometer (model 60/ED) with a Sodium D1 (yel-low) light source at Faculdade de Ciências da Uni-versidade de Lisboa. These measurements wereconducted at a controlled temperature (20±0.1◦C),using a thermostatic bath. This procedure was re-peated for all the fluids under study following theuser’s manual instructions.

The surfaces used for impact were character-ized in terms of wettability for all the liquids used.The static contact angle was obtained from an av-erage of 5 measurements in 5 different locationsof the stainless-steel surface, using an optical ten-siometer THETA (from Attension) with the sessiledrop method. These measurements are also rep-resented in Table 1.

3.3. Methodologies and post-processing3.3.1 Spray cone angle - SCA measurements

The high-speed camera was used to record thespray morphology. The videos were captured at13 029frames/second, with a resolution of 192 ×192pixel2 and using an exposure time of 10µm.

A maximum and minimum spatial resolutions of54.5µm/pixel and 70.6µm/pixel were obtained.

After this procedure, the recordings were ex-ported to 8 bits greyscale jpeg images. Then,these images are processed and analyzed by aMATLAB routine to compute the spray cone angle(SCA). The first step regards the image contrastenhancement and then the image background isremoved to reduce image noise (fig. 3(b)). After-wards, the interest zone is manually selected anda greyscale threshold is applied, converting the im-age into a binary one (fig. 3(c). The final step cre-ates a matrix with the left and right coordinates foreach Zi, that will be used as an input on the SCAexpression. The fig. 3(d)) is just an overlap of theoriginal image with the final.

(a) Step 1 (b) Step 2 (c) Step 3-4

(d) Over-lap

Figure 3: Spray cone angle: image processing steps

Thereafter, with the coordinates of each bound-ary, the spray cone angle is calculated followingEquation (5), that is based on triangles similarity:

SCA(◦) = 2 tan−1(x1 − xi

2Zi

)(5)

where x1 is the horizontal pixel difference betweenthe left and right boundaries at the 1st pixel of thezone of interest; xi is homologous to x1, but at theith pixel; Zi is the vertical difference between the1st and the ith pixel.

Figure 4: Spray cone angle scheme

It should be noted that the zone of interest for theSCA calculation corresponds to the stabilized zoneof the spray liquid sheet (before the liquid sheetbreakup). In these conditions, 90 different anglesare measured for each video frame, where the firstangle is measured 10 pixels under Z1, to decreasethe maximum absolute error. After the computationof those angles, the measurements for one videoframe are averaged. This process is repeated forthe rest of the video frames. The SCA measure-ments are represented in Table 1, with their respec-tive standard deviations.

4

Table 1: Thermophysical and wettability properties

Element DI Water Base Fluid Gold Silver

Geometry - - Spheres Cylinders Triangles

Mean Dimensions [nm] - - Diameter: 80Diameter: 12

Length: 39

Length: 30

Width: 17

Nanoparticle [wt.%] - - 0.1 0.5 1 0.1 0.5

Surfactant [wt.%] - 0.05 0.5 0.05 0.05 0.5 0.05

Dynamic Viscosity [cP ] 1.002 1.003 1.022 1.038 1.041 1.135 1.038

Specific Mass [Kg/m3] 998.21 998.71 999.16 1002.94 1007.68 999.16 1002.73

Surface Tension [mN/m] 76.34 34.29 33.11 37.75 37.72 32.67 37.30

Specific Heat [KJ/(KgK)] 4.18 4.18* 4.18 4.16 4.14 4.18 4.16

Thermal Conductivity [W/(mK)] 0.60 NA 0.61 0.65 0.70 0.61 0.65

Static Contact Angle [o] 97.55 62.81 53.38 58.10 58.92 51.52 58.77

Spray Cone Angle [o] @87psi 73.64 73.45 73.26 74.40 75.28 73.62 74.97

Regarding the uncertainties, they were only at-tributed to the uncertainty of the spray boundaries.Hence, a maximum error associated with their lo-cation was defined to be ±2 pixel. In this regard, amaximum error of 8.43◦ was computed.

3.3.2 Phase Doppler Anemometer measure-ments

Droplet size and velocity distributions were mea-sured using a two-component Phase DopplerAnemometer. This system was used with the BSAFlow Software v5.10 and its optical configurationand validation parameters are summarized in Ta-ble 2. It should be noted that the particle andflow P80 processor performs 1D measurements.Therefore, the normal velocity component U wasevaluated, given its importance in the empirical im-pact correlations used to predict the droplet impactoutcomes.

Table 2: Phase Doppler optical configuration and validation pa-rameters

Transmitting optics Receiving optics

Laser Power 300-400 mW Scattering angle 69o

Beam wavelength 514.5 nm Focal length 500 mm

Beam Spacing 60 mm Processor parameters

Focal length 310 mm Spherical Validation 10%Frequency shift 40 MHz S/N Validation -3dB

The measurements were performed at a dis-tance of 10 mm and 20 mm, which are correlatedto the first and second atomization moments [8],from the atomizer. The measured points consist ofa radial grid, where r = 0 mm corresponds to theradial origin of the spray axis. Initially, two sets ofmeasurement grids were adopted: −20 mm < r <20mm (for Z = 20mm) and−12mm < r < 12mm

(for Z = 10 mm) in 2 mm steps for two perpendic-ular axes, as shown in fig. 5.

-12mm

-20mm

+12mm

+20mm

Z

10mm

20mm

Figure 5: Measurement grid used with the phase Doppler sys-tem and coordinate system used.

The droplet size and velocity measurementswere exported from the BSA Flow Software andwere processed using a MATLAB routine. More-over, the referred distributions were characterizedby the arithmetic mean diameter (D10), defined as:

D10 =

∑Di∑Ni

(6)

where Di corresponds to the validated droplet di-ameter and Ni to the number of validated droplets.

3.3.3 Heat flux and wetted area measure-ments

Regarding the impact surface, an AISI 304 stain-less steel sheet with a predetermined area of90 × 60mm2, with a thickness of 20µm, was used.This surface was fixed to a metallic frame usinghigh-temperature silicone to reduce heat conduc-tion losses and to prevent electrical contact. Thebottom side of the surface was painted in blackwith an emissivity (ε) of 0.95. Then, Kapton tapewas applied on the surface edges to stretch it, andtwo rectangular copper electrodes were soldered

5

in opposite sides of the sheet. The final step wasto screw the metallic frame to the support, that col-lects and redirects the atomized liquid from the sur-face to the Reservoir 2, as represented in Figure 6.

(a) Top view (b) Bottom view

Figure 6: Metallic support with the stainless steel sheet surface

The high-speed thermographic camera Onca-MwIR-InSb-320 (Xenics) was used to evaluate theheat transfer at the surface bottom face. Thiscamera followed a custom-made calibration curvefor an iteration time of 200µs, defined in a pre-vious study [9] and converts Analog-Digital Unitsinto temperatures in Celsius. In this way, the ther-mographic camera recorded at 495frames/secondwith a resolution of 230 × 250 pixel2, that resultedin a spatial resolution of 222.22µm/pixel.

The high-speed camera recorded the spray im-pact, inferring important spray dynamics char-acteristics. Here, this camera was used at990frames/second, which corresponds to twotimes the frame rate of the thermographic cam-era recordings, with an exposure of 10µs. Theadopted resolution varied from the impact condi-tions: for Z = 20mm, a 520 × 320pixel2 resolutionwas used, whereas, for Z = 10mm, a resolutionof 520 × 256 pixel2 was adopted. In both cases, aspatial resolution of 64.49µm/pixel was obtained.

The spray cooling experiments were also per-formed at two different impingement heights, us-ing two different initial surface temperatures T0 ={74◦C; 145◦C}. It should be noted that each exper-imental condition was repeated at least 3 times. Asthe surface was heated by Joule effect, these tem-peratures correspond to the imposed currents ofI = {9A; 15A}, resulting in the imposed heat fluxesof q′′0= {855 W/m2; 2375 W/m2}.

The computation of the removed heat fluxes fromthe surface was made using a MATLAB script andit considers a 2-dimensional pixel by pixel energybalance. A detailed description is available in [9].From the tested conditions, a maximum uncertaintyof 0.55◦C, regarding the temperature resolution,was computed and it corresponds to a maximumrelative error of 2.70%.

An evaluation of the wetted area of the surfacewas also performed. It considers a threshold tem-

perature that is based on the maximum spatial tem-perature derivative across the horizontal coordi-nate. It is mainly located at the two-pixel interfacebetween the new wetted interface and the still driedarea and is computed by their mean temperature.

(a) Original image (b) Processed im-age

Figure 7: Wetted area computation demonstration

4. Results & discussion4.1. Nanoparticle effect on fluid propertiesThe nanofluids used in this study contain nanopar-ticles and are composed of different weight con-centrations, chemical element and shape. FromTable 1, it is clear that the nanoparticle addition in-duces small variations in the thermophysical prop-erties of the resulting nanofluids when comparedwith the base fluid.

Higher particle concentration leads, as ex-pected, to an increase in specific mass and ther-mal conductivity, whereas an opposite trend is ob-served for the specific heat. The thermal conduc-tivity expresses a major enhancement when com-pared with DI Water, showing an increase up to16.3% for the highest gold nanofluid concentration.These observations are in good agreement with[10, 11]. The addition of the nanoparticles has anegligible effect in the value of the refractive index.

The dynamic viscosity increases with the pres-ence of nanoparticles.This trend is in agreementwithin other studies (e.g. [6, 11]). When compar-ing the lowest concentration of gold nanofluids, onecan confirm that the cylindrically shaped nanopar-ticles, although smaller in size, induce a major in-crease on the dynamic viscosity when comparedwith the nanospheres.

It is detected that the surfactant plays a dominantrole in the surface tension value, deeply lowering it.From Table 1, the decrease of surface tension de-creases the static contact angle values, which en-hances the surface wettability. Nevertheless, whenadding nanoparticles, the surface tension slightlyincreases when compared to that of the base fluid,as opposed to the observations made in [4]. How-ever, between different nanoparticle materials andconcentrations, the surface tension values werehardly modified. This was also postulated in [8].

Regarding the SCA measurements, the additionof nanoparticles did not exert any influence outsidethe margin of uncertainty.

6

4.2. Droplet size and velocity analysis before impactAs the spray under study represented overall goodsymmetry and axis independence, the measure-ment grid was reduced to the positive radial po-sitions in one axis. In this order,the droplet meandiameter (D10) and the normal mean velocity com-ponent (U ) for Z= {10; 20} mm are represented inFigure 8.

0 4 8 1 2 1 6 2 02 53 03 54 04 55 05 56 0

Drop

let M

ean D

iamete

r (µm)

R a d i a l P o s i t i o n ( m m )0 4 8 1 2 1 6 2 02 5

3 03 54 04 55 05 56 0

Drop

let M

ean D

iamete

r (µm)

R a d i a l P o s i t i o n ( m m )

0 4 8 1 2 1 6 2 00

5

1 0

1 5

2 0

2 5

Axial

Mean

Veloc

ity (m

/s)

R a d i a l P o s i t i o n ( m m )0 4 8 1 2 1 6 2 00

5

1 0

1 5

2 0

2 5

Axial

Mean

Veloc

ity (m

/s)

R a d i a l P o s i t i o n ( m m )

Figure 8: Droplet mean characteristics measured with thePDA at different radial positions.The left column correspondsto Z=10mm, while the right column to Z=20mm.

The smallest and slowest droplets are found atnear centre of the spray (r= 0 mm) and are re-lated to the spray swirling motion and its hollowpattern. Therefore, these droplets are transportedto the centre due to the swirling motion, as theypossess a lower mass. For Z=10mm, as the radialdistance increases, the droplet diameter and veloc-ity increases until 6 mm < r < 8 mm. For r > 6mm,both values of D10 and U decrease as the dropletsstart to interact with the surrounding air mass at alower velocity.

Increasing the vertical distance (Z) to 20 mm, asimilar D10 and U behaviour is observable. How-ever, for this case, the interaction with the sur-rounding air is given at r = 8mm.

Performing a qualitative comparison between liq-uids with different surface tension, the decreaseof this value decreases the droplet mean diame-ter, which in turn decreases the axial velocity (U).These observations reinforce that surface tensionplays a major role in atomization (for the same vis-cosity value). Hence, the liquid sheet breakup gen-erates more droplets, since the energy needed toincrease the liquid surface area decreases.

The addition of nanoparticles to the base fluiddid not affect the values of diameter/velocity, withonly a few exceptions for Z=20mm (e.g. silver nan-otriangles and gold nanorods). It should be noted

that these exceptions can be attributed to experi-mental error or spray instabilities.

4.3. Impact predictionsFigure 9 represents the occurrence of the vari-ous droplet impact outcomes according to Bai etal. (2002) for a wetted surface at r=20mm andZ=20mm. From the previous subsection, the addi-tion of the surfactant slightly decreased the meandiameter and velocity, which in turn slightly de-crease the Reynolds number, and increase theWeber number, as surface tension decreased.These opposing trends will increase the Ohner-sorge number (Oh). Therefore, higher values of theOhnesorge number should correspond to slightlylower values of the Reynolds number, consideringisolated droplets. In this order, this should induce avertical displacement of the Oh-Re distribution forthose fluids relatively to DI Water. Hence, the out-come of droplet impact will slightly change to thesplash regime, as seen in fig. 9.

Regarding the effect of the addition of thenanoparticles, there are no observable changeswithin this range of concentrations when comparedto the base fluid. This observation is a good indica-tor that spray dynamics and impact outcomes areunaffected by nanoparticle presence in the baseliquid.

Figure 9: Impact outcomes prediction for the considered fluids(r=20mm; Z=20mm)

4.4. Heat transfer analysisFigure 10 represents relevant captured instancesduring wall impact using the IR camera and HScamera. IR images also show the temperaturedistribution across the radius, with an impact dis-tance of 10 mm and an initial surface temperatureof 145◦C.

The frame before liquid contact sets t0, wherethe initial temperature is at a constant temperature.At 2ms, the initial contact is made, and the spray isjust a thin liquid structure, far from developed. Att=8ms, although roughly visible, the spray starts toform a hollow liquid sheet and some disruptive lig-aments start to be visible. Hence, the formation ofthe first droplets, due to liquid sheet breakup anddue to impact, start to arise. At 28ms, an almost

7

closed hollow smooth liquid sheet, contracted bysurface tension forces, is formed. One can qualita-tively say that the droplets, resulting from splash orprimary atomization, become smaller.

0 5 1 0 1 5 2 002 04 06 08 0

1 0 01 2 01 4 01 6 0

0 5 1 0 1 5 2 002 04 06 08 0

1 0 01 2 01 4 01 6 0

0 5 1 0 1 5 2 002 04 06 08 0

1 0 01 2 01 4 01 6 0

0 5 1 0 1 5 2 002 04 06 08 0

1 0 01 2 01 4 01 6 0

0 5 1 0 1 5 2 002 04 06 08 0

1 0 01 2 01 4 01 6 0

0 5 1 0 1 5 2 002 04 06 08 0

1 0 01 2 01 4 01 6 0

2 0 0 m s

7 0 m s

2 8 m s

8 m s

R a d i a l D i s t a n c e ( m m ) 2 m sTemp

eratur

e local(º

C)

R a d i a l D i s t a n c e ( m m )

Temp

eratur

e local(º

C)Te

mpera

ture loc

al(ºC)

R a d i a l D i s t a n c e ( m m )

Temp

eratur

e local(º

C)

R a d i a l D i s t a n c e ( m m )

5 0 3 m s

Temp

eratur

e local(º

C)

R a d i a l D i s t a n c e ( m m )

Temp

eratur

e local(º

C)

R a d i a l D i s t a n c e ( m m )

Figure 10: a)Temperature variation along the radial profile(identified in the IR images by the black dashed line) duringspray impact on the heated surface, b) IR images taken fromthe backside of the surface, synchronized with c) Side views(taken with the high-speed camera) of the spray showing its dy-namics for DI Water. (q"=2375W/m2, Z = 10mm)

As the velocities increase, at 34ms, the surfacetension forces do not hold the closed bubble, as itstarts to straight the cone, becoming almost devel-oped. Also, at this point, a major part of the ana-lyzed surface area is wetted by the impacting liquid,forming a liquid film, due to deposition. Hence, theoverall removed heat fluxes become maximum, asthe active cooling area or film liquid line front be-comes higher. At later time intervals ( t > 70ms),as the temperatures and heat fluxes start to sta-bilize, the formation of secondary droplets at theedges of the spray intensifies.

4.4.1 Effect of the surface initial temperature

The cooling transient curves for two different im-posed heat fluxes are shown in Figure 11(a) andFigure 11(b), for q′′imp = 2375 W/m2 andq′′imp = 855 W/m2, resulting in initial temper-atures of 145◦C and 74◦C, respectively. An in-

crease of the initial surface temperature increasesthe removed heat flux, as expected. The main rea-son is given to the increased temperature differ-ence ∆T between the surface and liquid. How-ever, for T0 = 145◦C, nucleation was not observed,meaning that only single-phase heat transfer wasachieved.

0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 00

3 06 09 0

1 2 01 5 01 8 0 T w = 7 4 º C

T w = 1 4 5 º C

Heat

Flux (

kW/m

2 )

T i m e ( m s )0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0

2 04 06 08 0

1 0 01 2 01 4 01 6 0 T w = 7 4 º C

T w = 1 4 5 º C

T i m e ( m s )

Temp

eratur

e wall(ºC

)

Figure 11: Temporal variation of the heat flux (left) and walltemperature (right) for DI Water (Z=10mm)

4.4.2 Effect of impingement distance

From Figure 12(a), higher impingement distancespositively impact the temperature decrease of thesurface. This observation is explained by theincrease of the cooling area due to the naturaldroplet dispersion. This behaviour is explicit inFigure 12(b), showing a higher cooling area forZ=20mm.

0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 02 03 04 05 06 07 08 0

Z = 1 0 m m Z = 2 0 m m

Temp

eratur

e wall(ºC

)

T i m e ( m s )0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0

02468

1 01 21 41 6

Wette

d Area

(cm2 )

T i m e ( m s )

Z = 1 0 m m Z = 2 0 m m A r e a D i f f e r e n c e

Figure 12: Temporal variation of the wall temperature (left) andwetted cooling area (right) for DI Water

4.4.3 Effect of surface tension and nanoparti-cles

From Figure 13(a), the decrease of the surface ten-sion speeds up the spray maximum heat flux. Thisis mainly related to the increase of the wetted cool-ing area, fig. 13(b), indicating that a lower surfacetension diminishes spray development time inter-val. It should be noted that this promotes the sur-face temperature decrease, and it is consistent forfluids with similar surface tension values, as fig. 14depicts.

Regarding the nanoparticle addition, figs. 14and 15 show the transient cooling curves forthe fluids under study. From the three differentconcentrations of gold spherical nanoparticles, itwas expected that higher nanoparticle concen-trations should increase the cooling performance.

8

0 1 0 0 2 0 0 3 0 0 4 0 00

5 01 0 01 5 02 0 02 5 0

� = 3 4 . 2 9 [ m N / m ] � = 7 6 . 3 4 [ m N / m ]

Heat

Flux (

kW/m

2 )

T i m e ( m s )0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0

02468

1 01 21 41 6

Wette

d Area

(cm2 )

T i m e ( m s )

� = 3 4 . 2 9 [ m N / m ] � = 7 6 . 3 4 [ m N / m ] A r e a D i f f e r e n c e

Figure 13: Effect of surface tension on the: Space averagedheat flux (left) and wetted area (right) (T0=145◦C, Z=20mm)

However, the highest concentration performed theworst relative to the lower ones. The one with0.5 (wt.%) performed better, achieving a maximumheat flux of 295.9 kW/m2. Nevertheless, this valueis only 7.7% higher than the one for 0.1 (wt.%) andindicates that the concentration dependence is notsignificant within the 0.1 - 1 (wt.%) range.

0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 002 04 06 08 0

1 0 01 2 01 4 01 6 0

D I W a t e r B a s e F l u i dG o l d S p h e r i c a l

0 . 1 ( w t . % ) 0 . 5 ( w t . % ) 1 ( w t . % )G o l d C y l i n d r i c a l

0 . 1 ( w t . % )S i l v e r T r i a n g u l a r

0 . 5 ( w t . % )

Temp

eratur

e wall (º

C)

T i m e ( m s )Figure 14: Transient variation of the space averaged surfacetemperature

Comparing different geometries of gold nanopar-ticles at the same concentration, no differences areobserved. The same happens when changing thenanoparticle base material from gold to silver at 0.5(wt.%). Nevertheless, the comparable cooling per-formance between the two materials is consistentwith their similar thermal properties (table 1).

As Figure 15 shows, the nanoparticles increasedthe maximum removed heat flux from the wallat higher temperatures (from 110.8 − 116.6 ◦C),slightly enhancing cooling performance, at the ex-ception of gold nanospheres at 1(wt.%). Addition-ally, when computing the heat transfer coefficientas h = q′′w/ (Tw − Tf ) with Tf = 20 ◦C, one cancalculate the ratios of hnf/hBF or hnf/hDW for dif-ferent wall temperatures. The nanofluids showedan overall higher heat transfer coefficient whencompared to water and the base fluid (water andCTAB). Regarding their h value at q′′max, the use ofnanofluids resulted in enhancement from 9.8% to21.9% when compared to the base fluid and 11.5%to 38.8% when compared with DI water.

Performing a comparison between nanofluidsand DI Water, differences are evident. The heatfluxes are lower for DI Water, whose maximumvalue is given at a lower temperature (105 ◦C).Overall. the reason for these differences lye under

0 2 0 4 0 6 0 8 0 1 0 0 1 2 0 1 4 0 1 6 00

5 01 0 01 5 02 0 02 5 03 0 03 5 0

0

5 0

1 0 0

1 5 0

2 0 0

2 5 0

3 0 0

3 5 0q ' ' m a x ( k W / m 2 )

D I W a t e r 2 2 0 , 7 B a s e F l u i d 2 5 7 , 6

G o l d S p h e r i c a l0 . 1 w t ( % ) 2 7 4 . 00 . 5 w t ( % ) 2 9 5 . 9 1 w t ( % ) 2 1 7 . 7

G o l d C y l i n d r i c a l 0 . 1 w t ( % ) 2 8 5 . 8

S i l v e r T r i a n g u l a r 0 . 5 w t ( % ) 2 7 7 , 7

Heat

Flux (

kW/m

2 )

T e m p e r a t u r e w a l l ( º C )Figure 15: Variation of the overall space averaged heat fluxwith the space averaged surface temperature

two major dynamic changes:

• Spray dynamics – the lower surface ten-sion caused by the addition of the surfactantpromotes the liquid sheet breakup, promot-ing droplet formation and the occurrence ofq”max;

• Liquid thermal dynamics – the surfactant in-creases the surface wettability that promotesthe liquid-surface contact, and the nanoparti-cle addition improves the liquid’s thermal con-duction.

In this way, the superposition of these two ex-isting dynamics is responsible for the higher cool-ing performance of the base fluid and consequentlyof the nanofluids, as opposed to the distilled wa-ter. These mechanisms will translate to lower heattransfer coefficients at higher temperatures or onthe initial contact with the surface for the DI waterspray.

5. ConclusionsThe study carried out through this dissertationaimed at describing the heat transfer phenomenathat takes place in spray cooling using a new com-bination of diagnostic techniques. To this end, sev-eral nanofluids were atomized, taking advantage oftheir superior thermodynamic properties, inferringthe effects of changing the nanoparticle chemicalelement, geometry and concentration on the spraydynamics and cooling performance.

Using the phase Doppler anemometer, it wasverified that surface tension plays a dominant rolein the atomization process. Its value decreaseenhances the Weber number that promoted thedroplet formation rate, which in turn reduces theoverall droplet size. The addition of nanoparticlesin the base fluid did not modify the atomization pro-cess nor the impact outcomes within this concen-tration range. This is a positive indicator, mean-ing that any differences regarding the cooling per-formance are mainly attributed to heat transfer dy-namics at the liquid-solid surface interface.

Afterwards, the experimental installation wasadapted to accommodate a heated thin stainless-

9

steel surface. A high-speed time-resolved infraredcamera was placed beneath it to detail the heattransfer and to explore its use to describe thesephenomena. Additionally, this camera was syn-chronized with a high-speed camera that capturedlateral images of the spray impacting the surface.This synchronization was helpful to verify the sprayactuation repeatability and to examine the differentheat transfer mechanisms, although absent. Ac-cording to the results, this method proved to be re-liable as it captured details of the temperature vari-ations from spray impingement. It should also bepointed out that the results were repeatable, mainlyattributed to the use of the solenoid valve.

As the surface-to-nozzle distance increasedfrom 10 to 20 mm, the heat flux increased. Thisis mainly attributed to the increased area thatthe spray can cover at a higher surface to nozzledistance. Regarding the initial temperature influ-ence, as heat transfer was mainly given throughsingle-phase heat transfer, the removed heat fluxincreased as the initial temperature increased, asexpected. Moreover, the lower surface tensionenhanced the transient evolution of the wettedarea. This is associated with the decreased dropletcontact angle during impact and with the atomiza-tion enhancement, which promotes the maximumremoved heat flux. Regarding the addition ofnanoparticles, within the tested concentrations,no major changes were observed. However,the lowest concentrations of gold nanospheresconcentration slightly enhanced the maximumaveraged heat flux, which in turn increased theheat transfer coefficient. The opposite was ob-tained with the highest concentration. Changingtheir geometry from spheres to cylinders, heattransfer did not express any differences in heattransfer. The same result was obtained for differ-ent nanoparticle elements. This shows that theseparameters do not influence the heat transfer forthis set of concentrations and conditions.

For future studies, it is suggested that the rangeof nanoparticle concentration should be increasedto cause considerate thermophysical changes. Im-provements should be made on the heated surfacethermal isolation to diminish heat losses, and theimposed heat flux should also increase to triggertwo-phase heat transfer mechanisms. It would alsobe interesting to analyze the surface before and af-ter the nanofluid spray impact to check for nanopar-ticle deposition.

References

[1] Jungho Kim. Spray cooling heat transfer: Thestate of the art. International Journal of Heatand Fluid Flow, 28(4):753–767, 2007.

[2] Gangtao Liang and Issam Mudawar. Reviewof spray cooling–part 1: Single-phase and nu-cleate boiling regimes, and critical heat flux.International Journal of Heat and Mass Trans-fer, 115:1174–1205, 2017.

[3] Ruey-Hung Chen, Louis C Chow, and Jose ENavedo. Effects of spray characteristics oncritical heat flux in subcooled water spraycooling. International Journal of Heat andMass Transfer, 45(19):4033–4043, 2002.

[4] Shou-Shing Hsieh, Hsin-Yuan Leu, and Hao-Hsiang Liu. Spray cooling characteristicsof nanofluids for electronic power devices.Nanoscale research letters, 10(1):139, 2015.

[5] SMS Murshed, KC Leong, and C Yang. Ther-mophysical and electrokinetic properties ofnanofluids–a critical review. Applied ThermalEngineering, 28(17-18):2109–2125, 2008.

[6] Munish Gupta, Vinay Singh, Rajesh Kumar,and Z Said. A review on thermophysical prop-erties of nanofluids and heat transfer applica-tions. Renewable and Sustainable Energy Re-views, 74:638–670, 2017.

[7] Yanan Gan and Li Qiao. Evaporation char-acteristics of fuel droplets with the addition ofnanoparticles under natural and forced con-vections. International Journal of Heat andMass Transfer, 54(23-24):4913–4922, 2011.

[8] M Maly, AS Moita, J Jedelsky, APC Ribeiro,and ALN Moreira. Effect of nanoparti-cles concentration on the characteristics ofnanofluid sprays for cooling applications.Journal of Thermal Analysis and Calorimetry,135(6):3375–3386, 2019.

[9] Pedro Pontes. Thermographical analysis ofinterface heat transfer mechanisms, with hightemporal resolution. Master’s thesis, Univer-sidade de Lisboa Instituto Superior Técnico,2016.

[10] Shou-Shing Hsieh, Hao-Hsiang Liu, and Yi-Fan Yeh. Nanofluids spray heat transfer en-hancement. International Journal of Heat andMass Transfer, 94:104–118, 2016.

[11] Ji-Hwan Lee, Kyo Sik Hwang, Seok Pil Jang,Byeong Ho Lee, Jun Ho Kim, Stephen USChoi, and Chul Jin Choi. Effective viscosi-ties and thermal conductivities of aqueousnanofluids containing low volume concentra-tions of al2o3 nanoparticles. InternationalJournal of Heat and Mass Transfer, 51(11-12):2651–2656, 2008.

10