at SciVerse ScienceDirect

Applied Thermal Engineering 51 (2013) 839e851

Contents lists available

Applied Thermal Engineering

journal homepage: www.elsevier .com/locate/apthermeng

Influence of nanoparticle shape factor on convective heat transferand energetic performance of water-based SiO2 and ZnO nanofluids

Sébastien Ferrouillat a,*, André Bontemps a, Olivier Poncelet b, Olivier Soriano b,Jean-Antoine Gruss b

aUniversité Joseph Fourier, LEGI, BP 53, 38041 Grenoble, Cedex, FrancebCEA/LITEN/DTS/LETh, 17 avenue des martyrs, 38052 Grenoble, Cedex, France

h i g h l i g h t s

< We study convective heat transfer of four different water-based nanofluids.< ZnO and SiO2 nanoparticles each with two shape factors were used.< Thermal conductivities and viscosities have been measured.< Classical correlations are valid if measured thermal properties are used.< An energy performance evaluation criterion does not show significant improvement.

a r t i c l e i n f o

Article history:Received 9 February 2012Accepted 14 October 2012Available online 22 October 2012

Keywords:NanofluidSiO2

ZnOShape factorConvective heat transferPerformance energetic criterion

* Corresponding author. Tel.: þ33 (0)4 38 78 64 43E-mail address: sebastien.ferrouillat@ujf-grenoble

1359-4311/$ e see front matter � 2012 Elsevier Ltd.http://dx.doi.org/10.1016/j.applthermaleng.2012.10.02

a b s t r a c t

To appreciate the merits, in terms of energy, of two nanofluids and of two shapes of nanoparticles, anexperimental study has been carried out on water-based SiO2 and ZnO nanofluids flowing insidea horizontal tube whose wall temperature is imposed. Pressure drop and heat transfer coefficients havebeen measured at two different inlet temperatures (20 �C, 50 �C) in heating and/or cooling conditions atvarious flow rates (200< Re< 15,000). The Reynolds and Nusselt numbers have been determined byusing thermal conductivity and viscosity measured in the same conditions as those in tests. The resultsobtained show a small improvement of Nusselt numbers of studied nanofluids compared to those of thebase fluid. An energy Performance Evaluation Criterion (PEC) has been defined to compare heat transferrate to pumping power. Only nanofluid with ZnO nanoparticles having a shape factor greater than 3appears to reach a PEC as high as that of water.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

The need for convective heat transfer enhancement hasproduced a considerable amount of research. This enhancementcan be achieved by either active or passive ways. In this last caseconvective heat transfer can be improved by changing flowgeometry, boundary conditions, or by modifying thermophysicalproperties of the fluid. Enhancing thermal conductivity allows heattransfer coefficient to be increased. Among earlier efforts forenhancement of thermal conductivity the use of additives to liquidshas been explored [1] and more recently, the introduction of solidmicroparticles or nanoparticles in a base fluid [2]. These engineeredfluids are commonly named “nanofluids”. Recent studies have alsoshown that, if a thermal conductivity enhancement is possible byintroducing nanoparticles in a base fluid it is necessary to add

; fax: þ33 (0)4 38 78 51 61..fr (S. Ferrouillat).

All rights reserved.0

chemicals to stabilize the colloidal suspension which modify ther-mophysical properties of the solution. Moreover, an increase ofviscosity unfortunately occurs, leading to an extra pressure drop.Consequently, any gain in heat transfer could be compromised byan increase of pumping power.

To develop a nanofluid for heat transfer purposes, it is necessaryto have a global approach to take into account not only the thermalconductivity enhancement but also the modification of otherthermophysical properties. These properties determined with thehelp of models can have erroneous values and must be experi-mentally measured. On the other hand, it has been shown that, fora given concentration, nanoparticle shape plays an important roleboth in thermal conductivity and viscosity modification [3].

In this paper, we studied the effect of nanoparticle shape onpressure drop and heat transfer coefficient for water-based SiO2and ZnO nanofluids (2e5% wt) flowing in a horizontal pipe whosewall temperaturewas imposed. Moreover, these nanoparticles have

Nomenclature

Cp specific heat capacity, J kg�1 K�1

d tube diameter, mdh hydraulic diameter, mk thermal conductivity, Wm�1 K�1

L tube length, mM mass, kg_m mass flow rate, kg s�1

Nu Nusselt numberPr Prandtl number_Q heat flow rate, WR thermal resistance, �CW�1

Re Reynolds numberS heat exchange area, m2

Sp cross-sectional area, m2

T temperature, K, �CU overall heat transfer coefficient, Wm�2 K�1

V volume, m3

Greek symbolsa heat transfer coefficient, Wm�2 K�1

b shape factorDP pressure drop, PaDTlm log mean temperature difference, K4 nanofluid volume fraction4w nanofluid mass fractionL Darcy coefficientm dynamic viscosity, Pa sr density, km�3

j sphericity

Subscriptsb bulke externalexp experimentalf base fluidi internalin inletnf nanofluidout outlets nanoparticlet thermocouplew wall

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851840

also been chosen because the SiO2 thermal conductivity is ratherlow (1.3 or 1.4 Wm�1 K�1 following the crystalline state) and not sodifferent to that of water whereas the thermal conductivity of ZnOis rather high (60e110Wm�1 K�1) [4]. For the nanofluid itself someinteraction can exist between the base fluid and the nanoparticlemodifying average physical properties. This is the reason whythermophysical properties of these nanofluids have been experi-mentally determined. Then we have studied the effects of fluidcooling and fluid heating in forced convective heat transfer and theresults obtained have been compared with standard correlations.To appreciate themerits of nanofluids we have first considered heattransfer merits of the nanofluids by comparing measured Nusseltnumbers to those of the base fluid and then the merits from thepoint of view of energy in defining an energy Performance Evalu-ation Criterion which allows us to compare heat transfer rate topumping power.

2. Nanofluid preparation and characterization

2.1. Nanofluid preparation

The choice of nanofluids is the first key step in heat transferstudies. To evaluate thermal conductivity of particleefluidmixturesnumerous theoretical studies have been conducted dating back tothe classic work ofMaxwell [5]. He developed amodel to determinethe effective thermal conductivity for different volumetric loadingof spherical particles embedded in a base medium. This model hasbeen extended by Hamilton and Crosser [6] to non-sphericalparticles by introducing a shape factor b given by b¼ 3/j, wherej is the particle sphericity, defined as the ratio of the surface area ofa sphere with the same volume as that of the particle and thesurface area of the particle. The effective conductivity is expressedas follows:

k ¼ kfks þ ðb� 1Þkf � ðb� 1Þ�kf � ks

�4

ks þ ðb� 1Þkf þ�kf � ks

�4

(1)

where f is the volume fraction of the nanofluid; kf is the thermalconductivity of the base fluid; and ks is the thermal conductivity ofthe nanoparticles

The Maxwell formula corresponds to sphericity equals one.Several authors have proposed other models to take into accounteither the effects of the interface between the nanoparticle and thebase fluid or severalmicro-convection phenomena. Theywill not bediscussed here.

For b> 3, the thermal conductivity value is greater than the oneof spherical particles and increases with b. However, Timofeevaet al., [3] have shown that this enhancement is less than the onepredicted and they assume that it is due to interfacial thermalresistance. Nevertheless, they have observed an enhancement andthe greatest is for particles with a cylinder shape. We have chosento study nanofluids whose nanoparticles are small cylindricalshaped rods. Their performances have been compared withperformances of nanofluids whose particles are spherical or quasi-spherical and made of the same materials.

2.1.1. Water/silica nanofluidNanoparticles were produced in the LCNS laboratory of CEA

Grenoble. They are made of porous silica coated with silver.Organo-functionalised mesostructured silica nanoparticles wereprepared by surfactant templating (CTAB: cetyl ammoniumbromide) under alkaline conditions using a quenching techniquedescribed in details elsewhere [7].

Different molar percentages of MPTES (3-MercaptoPropyl-TriEthoxySilane) and TEOS (TetraEthyle OrthoSilicate) were co-condensed at a constant reaction molar ratio of 0.12 CTAB:0.5NaOH:1 SiO2:130 H2O. TheMPTES/TEOSmolar ratio in the synthesismixturewas less than 20mol%. In a typical reaction, CTAB (0.8 g)wasdissolved in an alkaline solution consisting of 10 g of 1 M NaOH in35.1 gH2O; 3.40 g and0.4 gofMPTES (10mol%)weremixed togetherand then added to the surfactant solution while stirring. After 40 s(dilution delay), the solution was added to 300 g of water whilestirring, which reduced the pH from 13 to approximately 12. Neu-tralisationof thismixture topH7was thenundertakenaftera further10 min (neutralisation delay) using 2 M HCl. Stirring was stoppedwhen the pH of the mixture had decreased to 7.0. By followingstrictly this recipe and by tuning the ratio MPTES/TEOS, it has beenpossible to design both rod-shaped nanoparticles and nanospheres.

Solids samples were obtained from the colloidal suspensions byevaporation in air for 24 h. Surfactant extraction was carried out

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851 841

using a Soxhlet apparatus with anhydrous ethanol as solvent.Finally FT-IR spectra were performed on the powders to check thecomplete loss of CTAB.

The powders were dispersed in a water solution AgNO3 0.1 M,then 0.1 M hydrated hydrazine water solution was added. Finallythe colloidal suspensions were dialysed in cellulosic membranes(MWCO: 14,000 Da) for 1 week against deionised water. The effi-ciency of the dialysis step was monitored by electrical conductivitymeasurements of the buffer water. Finally the solid percentages ofnanofluids were adjusted by evaporation or dilution. The nano-particles were characterized by transmission electron microscopy(TEM) performed on a JEOL 2000FX (Fig. 1). It appears that rod-shaped nanoparticles are rather “banana” shaped with a shapefactor comprised between 4 and 5.

Due to the fabrication protocol the actual concentration isdetermined at the end of the process. The mass concentration wasfound to be 2.4 % for sphere and 5 % for “banana” nanoparticles(volume concentration: 1.08% and 2.28%, respectively).

2.1.2. Water/zinc oxide nanofluidZnO aqueous colloidal suspensions both polygonal and rod-like

nanoparticles came from commercial sources, respectively, fromNyacol (SN15ES) and Evonik (VP DISP ZnO 20 DW). The colloidalsuspensions were dialyzed in cellulosic membranes (MWCO:14,000 Da) for 1 week against deionised water in order to removeall the organics and salts. The efficiency of the dialysis step wasmonitored by conductivity measurements of the buffer water.Finally the solid percentages of nanoparticles in the nanofluidswere adjusted by evaporation or dilution. The nanoparticles werecharacterized by transmission electron microscopy (TEM) per-formed on a JEOL 2000FX or on a High Resolution ScanningTransmission Electron Microscope (HRSTEM) Titan.

Contrary to the SiO2 nanoparticles it appears that the polygonalnanoparticles are massive as observed by means of HRSREM. It wasalso observed that they are perfectly crystalline (Fig. 2(a) and (b)).

The mass concentration was 4.4% for Nyacol� and 5% forEvonik� based suspensions (volume concentration: 0.82% and 0.93,respectively).

For both nanofluids, the surfactant effect is a consequence of thepreparation mode of the nanoparticles. A monolayer of repellentmolecules is formed at the particle surface. Due to the thinness ofthe layer, its effect is negligible on r and Cp values. Nevertheless themass density and the specific heat capacity have been experi-mentally verified.

Fig. 1. TEM photographs of SiO2 (Ag) nanoparticles: (a) spherical nan

For SiO2(Ag) particles, hydrated hydrazine is used to reducesilver salts (silver nitrate), the formed silver d0 metal nanoparticleswill interact with free thiol groups which wrap the silica nano-particles and so will form a continuous silver layer at the surface ofthe silica nanoparticles. Finally the colloidal silica sols are dialyzedagainst water to remove all the water-soluble by-products.

Concerning the commercial water/zinc oxide nanofluids aredialyzed (MWCO 14KD-30KD) against water to remove all thewater-soluble by-products (biocides, salts, surfactant).

So in the both cases, we can assume that the nanofluids areclean and that the influence of the remaining organics or salt on thethermal properties is poor.

2.2. Nanofluid characterisation

2.2.1. General2.2.1.1. Density. The density of the nanofluid is evaluated accordingto the standard formula:

r ¼ ð1� 4Þrf þ 4rs (2)

where 4 is the volume fraction of the nanofluid; rf is the density ofthe base fluid; and rs is the density of the nanoparticles.

For the use of nanofluids, this equation has been proved throughan experimental validation by Vaijiha et al. [8]. In our case, it hasbeen experimentally determined by weighing 100 ml of fresh silicaand zinc oxide suspensions. Uncertainty was calculated in propa-gating all precision uncertainties assuming a normal distribution.Measuring the mass M and the corresponding volume V, the massdensity is:

r ¼ MV

(3)

and the uncertainty is given by:

Dr ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�vr

vMdM

�2

þ�vr

vVdV

�2s

(4)

Measurement accuracywas between 0.1 and 0.5% and the valuesdeduced from Eq. (2) are within this range.

2.2.1.2. Specific heat. The formula of the specific heat for a mixtureis given by,

oparticles; and (b) nanoparticles with a shape factor (bananas).

Fig. 2. (a) HRSTEM image of ZnO polygonal nanoparticles (Nyacol�). (b) TEM image of ZnO nanoparticles with a shape factor (Evonik�).

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851842

Cp ¼ ð1� 4wÞCpfþ 4wCps

(5)

4w is themass fraction of the nanoparticle; Cpfis the specific heat of

the base fluid; and Cpsis the specific heat of the nanoparticles.

It has been also found appropriate for use with nanofluids asvalidated by O’Hanley et al. [9].

The obtained values are in accordance with measurementscarried out with a SETARAM micro calorimeter in the [20 �C, 70 �C]interval. For Cp measurements, the used DSC apparatus directlydeduces the Cp value. To determine the accuracy of the device wehave compared experimental values obtained with pure water andthose given by usual recommended values (VDI Heat Atlas, [10]).Relative error was found between 1 and 1.5%.

2.2.2. Thermal conductivity2.2.2.1. Water/silica nanofluid. Due to the lack of reliable models,especially for suspensions with non-spherical nanoparticles,thermal conductivity and dynamic viscosity have been measured.Thermal conductivities have been measured using a thermalproperty analyzer (model Lambda system 1, F5 Technology GmbH)based on the transient hot wiremethod. The accuracy was carefullychecked with pure water. The relative error of the thermalconductivity given by the manufacturer is 0.5%. However ,we havecompared experimental values obtained with pure water and thosegiven by usual recommended values [10] between 20 and 80 �C.Wehave found a relative difference of 1% below 70 �C and up to 2%above. Error bars of 2% are smaller than point sizes in Fig. 3.

Results are presented in Fig. 3(a) as a function of temperature.Two points can be underlined: (i) thermal conductivity of nanofluidswith spherical nanoparticles is less than that of water. This can beexplained by considering that the SiO2 nanoparticles are made ofaporousmaterial and that thenanoparticle conductivity isnot thatofpure silica. (ii) It is observed that thermal conductivity of nanofluidwith non-spherical nanoparticles appears to be slightly higher thanthat of spherical particles. However, the increase is not as high as theonecalculated fromtheHamiltoneCrosser relationship.Thebanana-shaped nanoparticles have a surface area in contact with stabilizingchemicals greater than the spherical ones leading to largeramountofthese chemicals inside the porous core. On the other hand, theinterfacial resistance is greater forbanana-likenanoparticles than forthe spherical onesand thendiminishes theeffect related to the shapefactor. The effect of interaction of nanoparticles with base fluids hasalso been reported by Timofeeva et al. [3].

2.2.2.2. Water/ZnO nanofluid. Thermal conductivity measurementsof ZnO suspensions have been carried out with the same analyzer

as for SiO2 suspensions. However, first results with rod-like nano-particles were not reproducible. It seems that such behaviour wasdue to the electrostatic behaviour of ZnO rods and special care wastaken in applying the hot wire method. As shown by Gautam [11],the two opposite ends of rod-shaped ZnO crystals can presentopposite electrostatic charges which lead to aggregates. This canfacilitate wire fouling. With the original hot wire, ZnO nanoparticledeposition was observed on the wire surface. It is the reason whythe hot wire was electrically insulated to limit polarisation andconsequently no wire fouling was detected.

Results are presented in Fig. 3(b) as a function of temperature. Itis observed that the thermal conductivity of the nanofluid isslightly greater than the water conductivity. The particles beingcrystallised no porosity effect occurs as for SiO2(Ag). Thermalconductivity increases more slowly than that of water withtemperature. This could be due to the thermal conductivity of ZnOwhich decreases with temperature [4] and compensates for waterconductivity augmentation. However, the enhancement is lowerthan that predicted by the HamiltoneCrosser relationship dueprobably to interface effects. It is also lower than that observedwithZnO/ethyleneglycol suspensions [12]

2.2.3. Dynamic viscosityDynamic viscosity was measured using a Brookfield rotational-

type viscometer. It is difficult to appreciate the viscosity uncer-tainty. First, experiments were carried out with a torque>500 nNm (the lowest limit of validity with this rheometer being10 nNm). Second, used viscosity values were those between shearrates between 100 and 1000 s�1. Within this range, viscosity valuesare constant and the fluid can be considered as Newtonian. Third,experiments were repeated several times and the found values arethe samewithin 4% but this is not a true statistical study. The resultsobtained as a function of temperature are presented Fig. 4 for1000 s�1 shear rate. For SiO2(Ag)/water suspensions, viscosity ofnanofluid with banana-like particles is close to the one withspherical nanoparticles. For ZnO/water suspensions, viscosity ofnanofluids with rod-shaped nanoparticles is slightly less than thatwith polygonal particles (Fig. 4).

For all nanofluids, as expected, viscosity is greater than that ofpure water.

3. Experimental apparatus and data reduction

3.1. Test loop and test section

The detailed description of the experimental apparatus hasalready been done previously [13]. The main features are recalled

0

100

200

300

400

500

600

700

800

0 10 20 30 40 50 60 70 80

Temperature (°C)

Th

erm

al C

on

du

ctivity (m

W.m

-1

.K

-1

)

Water

Silica sphere

Silica banana

Silica sphere predicted with Hamilton & Crosser

Silica banana predicted with Hamilton & Crosser

0

100

200

300

400

500

600

700

800

0 10 20 30 40 50 60 70 80

Temperature (°C)

Th

erm

al co

nd

uctiv

ity (m

W.m

.K)

WaterZnO Nyacol®ZnO Evonik®ZnO Nyacol® predicted with Hamilton & Crosser ZnO Evonik® predicted with Hamilton & Crosser

a

b

Fig. 3. Thermal conductivity of nanofluids. (a) Water/SiO2 (Ag) suspensions and (b) water/ZnO suspensions.

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851 843

hereafter. Flow loop used for pressure drop and convective heattransfer coefficient measurements with fixed wall temperatureboundary conditions is shown schematically in Fig. 5. Froma reservoir tank, the nanofluid, with specified concentration, wascirculated using a gear pump (Micropump, Ismatec, 0e200 l h�1).Assuming that nanofluids are considered as homogeneous fluids,the flow rate was measured by a Coriolis flowmeter (Micro MotionELITE, CFM10) that was calibrated with �0.1% accuracy over therange of 0e80 kg h�1. The pressure drop was measured directly bythree differential strain-gauge pressure transducers operating overa range of 0e1620 kPa with uncertainty within 0.075% f.s., as cali-brated by the manufacturer (Rosemount). A pH meter (EutechInstruments) was inserted downstream of the test section to followthe nanofluid pH change with a maximum accuracy of 0.01.

The test section (Fig. 6) consisted of a 0.5 m long tube-in-tubeheat exchanger, the tested nanofluid flowing into the 4 mm diam-eter and 1 mm thick inner copper tube (CuA1) and heating or

cooling water flowing into a 10 mm diameter and 1 mm thickstainless steel annular tube. The test section was preceded bya 0.5 m (125 diameters) adiabatic section.

The nanofluid was circulated inside the inner tube (primaryloop) with a temperature varying between 15 and 90 �C. To observethe potential influence of the transverse temperature gradient, thewater temperaturewas variedwithin the same range allowing us tochange the temperature difference between the fluid and the wall.The fluid could be heated or cooled thanks to various valves, andthen the gradient direction could be modified. After passingthrough the test section, the nanofluid entered a heat exchanger inwhich water was used as a cooling or heating fluid depending onnanofluid heating or cooling tests. For both primary and secondaryloops, temperature was controlled using two thermostatic baths(Polystat� 37, Fischer Scientific) and a second heat exchanger.

The entire test section was insulated with polyurethane foam(Armaflex) in order to minimize heat losses.

0.0000

0.0002

0.0004

0.0006

0.0008

0.0010

0.0012

0 10 20 30 40 50 60 70 80Température (°C)

Visco

sity (P

a.s)

Water Silica sphere Silica banana

0.0000

0.0002

0.0004

0.0006

0.0008

0.0010

0.0012

0 10 20 30 40 50 60 70 80

Temperature (°C)

Vis

co

sity

(P

a.s

)

Water ZnO Nyacol® ZnO Evonik®

a

b

Fig. 4. Dynamic viscosity of water/ZnO nanofluids. (a) Water/SiO2 (Ag) suspensions and (b) water/ZnO suspensions.

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851844

3.2. Temperature measurements

Two (K-type) thermocouples were inserted into the flow at theinlet and outlet of the test section for measuring bulk temperaturesof nanofluid. In order to increase the outlet temperature accuracyfor laminar flow, a static mixer was inserted downstream of the testsection. To record the temperature at the outer surface of thecopper tube and the bulk temperature, four (K-type) thermocou-ples were brazed on the inner tube wall and four (K-type) ther-mocouples were inserted into the inner tube at equally spaced10 cm distances (Fig. 6). The thermocouples were calibrated beforetests and had a maximum accuracy of �0.1 �C. All the data wererecorded by an Agilent 34970A data acquisition unit.

To determine inner wall temperature, the thermal resistancedue to conduction through the tube was taken into account. To

determine the inner flow bulk temperature Tbi we added a correc-tive term by taking into account the fin effect due to the intrusivethermocouples in writing an energy balance between forcedconvective flow perpendicular to the thermocouple and conductionin the thermocouple between its extremity and the wall [13].

3.3. Data reduction

The heat flow rate _Q was determined from the mass flow rate _mand the inlet and outlet temperatures of the fluid:

_Q ¼ _mCpðTin � ToutÞ (6)

The internal heat transfer coefficient ai between the nanofluidand the wall was derived from the following expression of the heatflow rate:

Pump

Gear pump

Nanofluidreservoir

Test sectionP

T

P Mixer

TPump

Acquisition System

T T

Water of city

Sewer

heatexchanger

Heated bath

T

Coriolis flow

TT T

T T T T

pH meter

Heated bath

Fig. 5. Schematic of convective loop experimental facility.

4 m m 6 m m 1 0 m m 1 2 m m

C o p p e r

C o p p e r

S t a i n l e s s s t e e l

S t a i n l e s s s t e e l

T W a l l ( K - t y p e t h e r m o c o u p l e )

T B u l k ( K - t y p e t h e r m o c o u p l e )

Fig. 6. Test section. Scheme of the thermocouple positioning.

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851 845

Fig. 7. Evolution of the Nusselt number accuracy as a function of Reynolds number for ZnO Nyacol� heating condition.

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851846

0B 1

1C

_Q ¼ B@1ai

þ Rw

CASðTwe � TbiÞ (7)

where S is the heat exchange area (m2), Twe is the average externalwall temperature of the four K-type thermocouples brazed on theinner tube (K), Tbi is the average internal bulk temperature deducedfrom the four K-type thermocouples inserted into the inner tube(K), and Rw is the thermal resistance of the copper tube wall(mKW�1).

This thermal resistance Rw is given by:

Rw ¼ di2kw

ln�dedi

�(8)

where: di and de are, respectively, the inner and outer diameters ofthe inner tube (m), kw is the thermal conductivity of the inner tube(Wm�1 K�1).

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0 2000 4000 6000 80

Reynolds

Darcy co

efficien

t

ZnO Evonik® heatingSilica banana heatingZnO Evonik® coolingSilica banana coolingWater heatingWater (Blasius and Poiseui

Fig. 8. Evolution of the Darcy coefficient as a function of Reynolds number. F

The internal heat transfer coefficient ai (Wm�2 K�1) can thus becalculated from

ai ¼�SðTwe � TbiÞ _Q

� Rw

��1(9)

Once the experimental heat transfer coefficient ai is determined,the experimental Nusselt number must be compared with thevalue obtained experimentally with pure water, which is the basefluid. This comparison is done by plotting the ratio of the Nusseltnumber measured with the nanofluid Nunf and the Nusselt numbermeasured with pure water Nuf. In each case, the Reynolds numberwas deduced from the mass flow rate measurement by:

Re ¼ 4 _mpdim

(10)

where m is the measured fluid dynamic viscosity taken at averagebulk temperature. Knowing the exact value of viscosity is crucial

00 10000 12000 14000 16000

number

ZnO Nyacol® heatingSilica sphere heatingZnO Nyacol® coolingSilica sphere coolingWater cooling

lle)

ull line refers to Poiseuille (Re< 2300) and Blasius (Re� 2300) relations.

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851 847

because incorrect determination of the Reynolds number can causea shift in the curves and lead to misinterpretation of the Nunf/Nufratio.

Using three differential strain-gauge pressure transducers, thepressure drop measurement enables the Darcy coefficient (or fric-tion factor) to be deduced with the following expression:

Lexp ¼ 2DPdhL

rS2pi_m2 (11)

The maximum relative uncertainties of the Reynolds numberand Darcy coefficient were estimated and are, respectively, lessthan 1.9% and 4.3%. The uncertainty on the Reynolds number wasestimated in taking the viscosity value obtained with the shear rateof 1000 s�1. For very low shear rates (low flow rates) the mode of

0.90

0.95

1.00

1.05

1.10

1.15

1.20

0 2000 4000 6000 8Reynold

Nu

nf / N

uf

Silica sphere heatingSilica sphere cooling

0.90

0.95

1.00

1.05

1.10

1.15

1.20

0 2000 4000 6000 8

Reynold

Nu

nf / N

uf

ZnO Nyacol® heatingZnO Nyacol® cooling

a

b

Fig. 9. Evolution of the measured ratio Nunf/Nuf as a function of Reynolds n

calculus cannot to be adapted due to the non-Newtonian behaviourof the fluid. The maximum relative uncertainty of the Nusseltnumber is highly dependent on the Reynolds number. For example,this uncertainty is between 78% and 5% for a Reynolds numberbetween 100 and 12,000. Details of uncertainty calculations aregiven in Ref. [13]. We can remark that, uncertainty on Nusseltnumbers is very high for low Reynolds numbers (100e1000). This isessentially due to the small temperature differences betweenthermocouples for very low flow rates. This is the reason why wehave limited the analysis to Reynolds corresponding to the transi-tion and the turbulent regimes (Re> 300). In this case the uncer-tainty is comprised between 11% and 5%. For example, Fig. 7 showsthe relative uncertainty on the Nusselt number calculated with thepropagation method for Nyacol in heating conditions. We canobserve the strong dependence on Reynolds number.

000 10000 12000 14000 16000s number

Silica banana heatingSilica banana cooling

000 10000 12000 14000 16000

s number

ZnO Evonik® heatingZnO Evonik® cooling

umber. (a) Water/SiO2 (Ag) suspensions and (b) water/ZnO suspensions.

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851848

4. Experimental results and discussion

4.1. Pressure drop

A preliminary test was conducted with water for pressure lossmeasurements. Three measurement conditions were studied. Thefirst (not presented here) is an isothermal condition, where the twofluids are at the same temperature (20 �C or 50 �C). The second isa cooling condition for which the external water is at 20 �C and thenanofluid at 50 �C. The third is a heating condition for which theexternal water is at 50 �C and the nanofluid at 20 �C.

The results obtained (Fig. 8) were compared with classicalrelationships. In laminar flow regime (Re< 2300), the followingPoiseuille equation is used in the calculations:

L ¼ 64Re

(12)

In turbulent flow regime, the Blasius equation is used:

L ¼ 0:316Re�0:25 (13)

In heat transfer conditions, the Poiseuille and the Blasius lawsare followed provided that the experimental Darcy coefficient ismodified by using a corrective factor as indicated by Petukhov [14]:

L ¼ Lexpðmw=mÞm (14)

where mw is the viscosity of the fluid near the wall and m is theviscosity at the bulk temperature. The m exponent was experi-mentally found to be equal to the following:

- for heating conditions, m¼�0.58 for laminar flow andm¼�0.25 in turbulent flow; and

- for cooling conditions, m¼�0.50 for laminar flow andm¼�0.25 for turbulent flow.

At first observation, in laminar flow regime, the results followedPoiseuille classical laws with demineralised water at Re< 1000. Asalready remarked, for very low Reynolds numbers the nanofluidscould not follow the Poiseuille law due to their non-Newtonianbehaviour. For 1000< Re< 2300, the Poiseuille law under-

0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40

Predicted Nu

Measu

red

N

usselt n

um

berr

Water heatingWater coolingZnO Nyacol® heatingZnO Evonik® heatingSilica sphere heatingSilica banana heatingZnO Nyacol® coolingZnO Evonik® coolingSilica sphere coolingSilica banana coolingGnielinski correlation10%-10%

Fig. 10. Comparison of measured Nusselt numbers and Nus

predicts the experimental values. This deviation may be associ-ated with the presence of the four thermocouples inserted into theinner tube. These thermocouples may generate turbulence fora Reynolds number lower than 2300 and induce an increasedpressure drop and thus an increase of the Darcy coefficient. Asa result, a smooth transition between laminar and turbulent regimeis observed compared to the sudden change of pressure drop at theturbulence onset.

The pressure drop of nanofluid with banana- or rod-like nano-particles is systematically smaller than the one with spherical orpolygonal particles. This could be due to the densities of thebanana- and rod-like nanoparticle suspensions which are greaterthan the one of spherical (resp. polygonal) nanoparticles.

4.2. Heat transfer coefficient and Nusselt number

Before determining the convective heat transfer coefficient ofa nanofluid, the apparatus was calibrated using pure demineralisedwater in heating and cooling conditions. Then, heat transfer coef-ficients of the two water/SiO2(Ag) nanofluids (with spherical andbanana-like nanoparticles) and of the two water/ZnO nanofluids(with polygonal and rod-like nanoparticles) were determined andNusselt numbers deduced. To compare this Nusselt number withthose of the basefluid, we have reported the ratio Nunf/Nuf whereNunf and Nuf are the Nusselt numbers for the nanofluid and water,respectively. In Fig. 9(a), it is observed that Nusselt numbers of SiO2nanofluids are mixed together and are about 4% greater than theone of pure water. It is difficult to distinguish whether or not thebehaviour of nanofluids with spherical nanoparticles is different tothe ones with nanoparticles having a shape factor. For ZnO nano-fluids (Fig. 9(B)), Nusselt number ratios are shared in two groups:one for polygonal nanoparticles suspensions (8% increase), theother for rod-like nanoparticles suspensions (3% increase). Thishigher augmentation for polygonal particles could be due to thedynamic viscosity of the rod-like ZnO suspensionwhich is less thanthe one of polygonal ZnO suspensions. As a simplistic approach,using the DittuseBoelter correlation [15], in turbulent regime foridentical Reynolds numbers the ratio Nunf/Nuf is function of m0:3nf .Considering viscosity values this ratio is higher for polygonalparticles than rod-like particles.

50 60 70 80 90 100

sselt number

selt numbers deduced from the Gnielinski correlation.

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851 849

These results have been interpreted by using the classicalGnielinski correlation valid for Re> 2300 in transition and turbu-lent regime in heating and cooling conditions [16] and given by:

Nu ¼ ðL=8ÞðRe� 1000ÞPr1þ 12:7

ffiffiffiffiffiffiffiffiffiffiffiffiffiðL=8Þp �Pr2=3 � 1

�� PrPrw

�0:11�1þ

�dhL

�2=3�m

(15)

In this formula, the Darcy coefficient is given byL¼ (1.82 log10Re� 1.64)�2, where Re is the Reynolds number, Prand Prw are the Prandtl numbers calculated at the fluid bulktemperature and at the inner wall temperature, respectively, L isthe tube length and dh the hydraulic diameter. The bulk tempera-ture is an average between the inlet and outlet fluid temperatures.

It canbe seen that theexperimentaldata correspondwellwith thepredictionsof thecorrelation towithin�10%(Fig.10). This conclusionhas already been drawn byWilliams et al. [17]. It must be noted thatGnielinski gives a �20% range of validity for his correlation.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 2000 4000 6000 80Reynold

PE

Cn

f/P

EC

f

Silica sphere heatingSilica sphere cooling

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 2000 4000 6000 80Reynolds

PE

Cn

f/P

EC

f

ZnO Nyacol® heatingZnO Nyacol® cooling

a

b

Fig. 11. Evolution of ratios of energetic performance ev

4.3. Optimisation and energy criteria

There are several ways to characterise the energy or thermalperformance of a fluid flowing in a specific device.

(a) The first one is to only consider the heat transfer coefficient orthe Nusselt number enhancement compared with a referenceone. In the preceding paragraph we have seen that for SiO2suspensions a slight enhancement is observed whereas noimprovement is brought with ZnO suspensions.

(b) As heat transfer and pressure drop are the most critical factors,they can be compared through several approaches: (i) the heattransfer per unit of pressure drop represented by the ratio of thej-Colburn factorby the f-friction factor (orDarcycoefficient) [18]or (ii) performance criteria allowing a device to be comparedwith a reference device by defining the following JF factor [19]:

JF ¼ j=jRðL=LRÞ1=3

(16)

00 10000 12000 14000 16000s number

Silica banana heatingSilica banana cooling

00 10000 12000 14000 16000 number

ZnO Evonik® heatingZnO Evonik® cooling

aluation criteria as a function of Reynolds number.

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851850

Parameters with the R subscript are those of the referencedevice. This JF factor was used to compare two heat exchangershaving a different geometry. It could be used to compare two heatexchangers with the same geometry but with two different fluids.

(c) Rather than the two preceding approaches we have preferredto use the energy PEC (Performance Evaluation Criterion)defined below and based on an energy global approach. It isdefined as the ratio of heat flow rate transferred to the requiredpumping power in the test section:

PEC ¼ _m$CpðTout � TinÞ_v$DP

(17)

where _m is the mass flow rate (kg/s), _v the volumic flow rate (m3/s),Tin and Tout are the tube inlet and outlet temperatures, repectively,and DP the pressure drop (Pa).

This criterion is directly related to gains and losses of energy inan industrial plant. Such a criterion has already been used inequivalent forms by different authors [20,21]. As for heat transferwhere Nusselt numbers have been comparedwith that of water, wehave plotted on the same figure the PECnf/PECf ratio for the fournanofluids (Fig. 11).

As previously observed with commercial SiO2 nanoparticles [13]the PEC values of studied nanofluids with SiO2(Ag) nanoparticlesare less than those of water whatever the shape factor indicatingthat the energy budget is unfavourable. For ZnO/water nanofluids,it appears that the PEC ratio values are close to unity for nano-particles with a shape factor.

4.4. Nanofluid stability

Commercial applications of nanofluids are limited due, in part,to the limited understanding of stability and rheology. Manyresearchers are currently studying the effects of nanofluidcomposition and experimental conditions on nanofluid stability.Ghadimi et al. [22] have listed the major factors influencing theenhancement of heat transfer: chemical composition of the solidparticle and the base fluid, particle source and concentration,particle shape and size, surfactants, temperature, pH value (surfacecharge), monodispersity, IsoElectricPoint and elapsed time. Inconvective heat transfer we can add experimental conditions suchas flow shear stress, temperature level and temperature gradients.Hwang et al. have shown the influence of the interaction betweenthe base fluid and the particles on thermal conductivity [23].Witharana et al. have studied the influence of the flow shear stresson stability [24]. For ZnO nanofluids, it has been shown by someauthors that stable suspensions were obtained for pH <7.2 and pH>12 [25] and by others in the pH range 8.58 and 9.38 showing theimportance of surfactant effect [26]. On the other hand, ZnOsuspensions subjected to heating and cooling cycles can presentdifferent viscosity variations following the base fluid nature (wateror waterþ surfactant). A strong hysteresis also was observed incooling conditions as was also found with Al2O3ewater nanofluids[27]. For SiO2 and Al2O3 suspensions, existence of a criticaltemperature has been revealed beyond which nanofluid propertiesare strongly modified [13,27]. All these results show that it isnecessary to measure thermal properties of nanofluids in the sameconditions as experimental ones. In our case, thermal conductivityhas been measured after a time sufficient to obtain reproducibleresults. This procedure seems to indicate that nanofluid modifica-tions occur during this time. Several authors have proposed thatparticle aggregation can provoke either enhancement or loweringof thermal conductivity [22] and the controversy is not over [28].

These contradictory behaviours have led us to measure thermalproperties as closely as possible to experimental conditions.

5. Conclusions

Experiments have been conducted to study the influence of theshape factor of nanoparticles in colloidal suspensions in order todetermine the energetic performance of these fluids. Two nano-particle materials have been used, SiO2 recovered with Ag and ZnO,both with water as base fluid. Spherical and “banana-like” particleswere used for SiO2/water nanofluids. Polygonal and rod-like werestudied for ZnO particle suspensions. Thermal conductivity hasbeen measured for the 4 nanofluids in conditions close to theexperimental ones. Dynamic viscosity as a function of temperaturehas also been measured. Convective heat transfer has been studiedfor the 4 nanofluids flowing inside a horizontal tube in cooling andheating conditions. From the measurements of pressure drop andheat transfer coefficients Darcy coefficients and Nusselt numbershave been deduced. Finally, the energetic performances have beencharacterized by using the PEC (Performance Evaluation Criterion).

From these measurements, the general conclusions can bedrawn.

- Stability of nanofluids must be achieved and verified.- Measurements of physical properties of nanofluids must becarried out in conditions close to the usage ones.

- If the measured thermal and physical properties are taken intoaccount to calculate the dimensionless numbers, the existingcorrelations reproduce the convective heat transfer and thepressure loss in tubes within their range of validity.

- From an energy point of view, it is difficult to obtain a PEChigher than that of water. It seems that with nanofluids withnanoparticles whose shape factor is greater than 3 this objec-tive could be reached even exceeded.

Further studies must be undertaken to engineer new nanofluidswhose physical properties are stable with time in industrialconditions. Combining material, size and shape factor as well asbase fluid it should be possible to fulfil the required specifications.

Acknowledgements

This work was partially supported by the “Programme Inter-disciplinaire Energie” of the CNRS (Centre National de la RechercheScientifique) in the frame of the “Microcond” project and by theEnvironment and Energy Management Agency (ADEME) undergrant N� 0566C00.

References

[1] A.E. Bergles, Survey and evaluation of techniques to augment convective heatand mass transfer, Prog. Heat Mass Transfer 1 (1969) 331e424.

[2] S.U.S. Choi, Enhancing thermal conductivity of fluids with nanoparticles, in:D.A. Siginer, H.P. Wang (Eds.), Development and Applications of Non-Newtonian Flows, vol. 231/66, ASME (FED/MD), New York, 1995, pp. 99e105.

[3] E.V. Timofeeva, J.L. Rotbort, D. Singh, Particle shape effects on thermophysicalproperties of alumina nanofluids, J. Appl. Phys. 106 (014304) (2009) 1e10.

[4] C.F. Klingshim, B.K. Meyer, A. Waag, A. Hoffmann, J. Geurts, Zinc Oxide. FromFundamental Properties Towards Novel Applications, Springer-Verlag, Berlin,Heidelberg, 2010.

[5] J.C. Maxwell, A Treatise on Electricity and Magnetism, 2nd ed., vol. 1, Clar-endon Press, Oxford, 1981.

[6] R.L. Hamilton, O.K. Crosser, Thermal conductivity of heterogeneous two-component systems, I&EC Fund 1 (3) (1962) 187e191.

[7] Sandra L. Burkett, Stephen D. Sims, Stephen Mann, Synthesis of hybrid inor-ganiceorganic mesoporous silica by co-condensation of siloxane and orga-nosiloxane precursors, Chem. Commun. 11 (1996) 1367e1368.

S. Ferrouillat et al. / Applied Thermal Engineering 51 (2013) 839e851 851

[8] R.S. Vaijha, D.K. Das, B.M. Mahagaonkar, Density measurement of differentnanofluids and their comparison with theory, Petrol. Sci. Technol. 27 (6)(2009) 612e624.

[9] H. O’Hanley, J. Buongiorno, T. McKrell, L.-W. Hu, Measurement and modelvalidation of specific heat capacity with differential scanning spectrometry,Adv Mech Eng (2012). Article ID 181079, 6 pp.

[10] VDI Heat Atlas, Springer e Verlag, Berlin, Heidelberg, 2010.[11] U. Gautam, First Observation of New Type of “Sea Urchin” Shaped Crystalline

Aggregate, International Center for Materials Nanoarchtechtonics (MANA),National Institute for Materials Science, Japan, 2010, NIMS News.

[12] W. Yu, H. Xie, L. Chen, Y. Li, Investigation of thermal conductivity andviscosity ofethyleneglycol basedZnOnanofluid, Thermochim.Acta 491 (1-2) (2009)92e96.

[13] S. Ferrouillat, A. Bontemps, J.-P. Ribeiro, J.-A. Gruss, O. Soriano, Hydraulic andheat transfer study of SiO2/water nanofluids in horizontal tubes withimposed wall temperature boundary conditions, Int. J. Heat Fluid Flow 32(2011) 424e439.

[14] B.S. Petukhov, Heat transfer and friction in turbulent pipe flow with variablephysical properties, in: T.F. Irvine, J.P. Hartnett (Eds.), Advances in HeatTransfer, vol. 6, 1970, pp. 503e564.

[15] F.W. Dittus, L.M.K. Boelter, Heat Transfer in Automobile Radiators of theTubular Type, vol. 2, University of California Publications of Engineering, 1930[(13) pp. 443e461].

[16] V. Gnielinski, New equations for heat and mass transfer in turbulent pipe andchannelflow (Translated fromGerman), Int. Chem. Eng. 16 (2) (1976) 359e368.

[17] W. Williams, J. Buongiorno, L.-H. Hu, Experimental investigation of turbulentconvective heat transfer and pressure loss of alumina/water and zirconia/water nanoparticle colloids (nanofluids) in horizontal tubes, Trans. ASME, J.Heat Transfer 130 (042412) (2008) 1e7.

[18] A.P. Colburn, A method of correlating forced convection heat transfer data anda comparison with fluid friction, Trans. AIChE 29 (1933) 174.

[19] J.-Y. Yun, K.-S. Lee, Influence of design parameters on the heat transfer andflow friction characteristics of the heat exchanger with slit fins, Int. J. HeatMass Transfer 43 (2000) 2529e2539.

[20] B.C. Pak, Y.I. Cho,Hydrodynamic andheat transfer study of dispersedfluidswithsubmicron metallic oxide particles, Exp. Heat Transfer 11 (1998) 151e170.

[21] T. Maré, S. Halelfadl, O. Sow, P. Estellé, S. Duret, F. Bazantet, Comparison of thethermal performances of two nanofluids at low temperature in a plate heatexchanger, Exp. Therm. Fluid Sci. 35 (2011) 1535e1543.

[22] A. Ghadimi, R. Saidur, H.S.C. Metseleer, A review of nanofluid stability prop-erties and characterization in stationary conditions, Int. J. Heat Mass Transfer54 (17e18) (2011) 4051e4068.

[23] Y. Hwang, J.K. Lee, C.H. Lee, Y.M. Jung, S.I. Cheong, C.G. Lee, et al., Stability andthermal conductivity characteristics of nanofluids, Thermochim. Acta 455(2007) 70e74.

[24] S. Whitarana, H. Chen, Y. Ding, Stability of nanofluids in quiescent and shearflow fields, Nanoscale Res. Lett. 6 (2011) 231. 6 p.

[25] A. Degen, M. Kosec, Effect of pH and impurities on the surface charge of zincoxide in aqueous solution, J. Eur. Ceram. Soc. 20 (2000) 667e673.

[26] U. Ojha, S. Das, S. Chakraborty, Stability, pH and viscosity relationship in zincoxide based nanofluids subject to heating and cooling, J. Mater. Sci. Eng. 4 (7)(2010) 24e29.

[27] C.T. Nguyen, F. Desgranges, N. Galanis, G. Roy, T. Maré, H. Angue Mintsa,Viscosity data for Al2O3-water; nanofluid hysteresis: is heat transfer enhance-ment using nanofluids reliable? Int. J. Therm. Sci. 47 (2) (2008) 103e111.

[28] S.M. Sohel Murshed, Correction and comment on “thermal conductance ofnanofluids: is the controversy over?”, J. Nanoparticle Res. 11 (2009) 511e512.