You are on page 1of 4

Available online at www.sciencedirect.

com

Fluid Phase Equilibria 260 (2007) 275278

The thermal conductivity of alumina nanoparticles dispersed in ethylene glycol


Michael P. Beck, Tongfan Sun, Amyn S. Teja
School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0100, United States Received 29 March 2007; received in revised form 5 July 2007; accepted 10 July 2007 Available online 17 July 2007

Abstract The thermal conductivities of several nanouids (dispersions of alumina nanoparticles in ethylene glycol) were measured at temperatures ranging from 298 to 411 K using a liquid metal transient hot wire apparatus. Our measurements span the widest range of temperatures that have been investigated to date for any nanouid. A maximum in the thermal conductivity versus temperature behavior was observed at all mass fractions of nanoparticles, closely following the behavior of the base uid (ethylene glycol). Our results conrm that additional temperature contributions inherent in Brownian motion models are not necessary to describe the temperature dependence of the thermal conductivity of nanouids. Our results also show that the effect of mass or volume fraction of nanoparticles on the thermal conductivity of nanouids can be correlated using the Hamilton and Crosser or Yu and Choi models with one adjustable parameter (the shape factor in the Hamilton and Crosser model, or the ordered liquid layer thickness in the Yu and Choi model). 2007 Elsevier B.V. All rights reserved.
Keywords: Thermal conductivity; Nanouids; Alumina nanoparticles; Ethylene glycol

1. Introduction The thermal conductivity of uids containing dispersed solid particles has been studied for more than one hundred years [1], and theories to describe the increase in thermal conductivity of these dispersions with volume fraction of the particles have been proposed by Maxwell [1], Hamilton and Crosser [2], and others. More recently, dispersions containing nanoparticles (or nanouids) have attracted attention because they apparently exhibit a greater thermal conductivity increase than that predicted by the Maxwell model [3]. Choi, Eastman, and their co-workers were some of the rst to study such anomalous behavior, when they reported a 40% enhancement in the thermal conductivity of ethylene glycol with the addition of 0.3% (v/v) Cu nanoparticles [4], and a 150% enhancement in the thermal conductivity of a synthetic oil with the addition of 1% (v/v) carbon nanotubes [5]. Others have shown that this enhancement increases linearly with the volume fraction of nanoparticles

Corresponding author. Tel.: +1 404 894 3098; fax. +1 404 894 2866. E-mail address: amyn.teja@chbe.gatech.edu (A.S. Teja).

in the case of CuO in ethylene glycol [6], SiC in water and ethylene glycol [7], and Al2 O3 in several liquids [8]. However, as noted in these and other studies [4,5], the Maxwell [1] and Hamilton and Crosser [2] models were not able to predict these greatly enhanced thermal conductivities because the models do not include any dependence on particle size. The enhancement has been variously attributed to an ordered liquid layer at the particle surface [9], and to Brownian motion that creates microconvection between the particles and liquid molecules [10,11]. However, literature data have failed to conclusively validate either mechanism for thermal conductivity enhancement. The relationship between the thermal conductivity of nanouids and temperature has also proved to be somewhat difcult to generalize. Das et al. [12] have reported a linear relationship between the thermal conductivity of aqueous and ethylene glycol-based nanouids and temperature, and more recently, Zhang et al. [13] have reported that the slope of the thermal conductivity versus temperature relationship for nanouids is linear and approximately the same as that of the base uid. However, these and other studies have considered data over limited temperature ranges so that it has proved difcult to draw

0378-3812/$ see front matter 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.uid.2007.07.034

276

M.P. Beck et al. / Fluid Phase Equilibria 260 (2007) 275278

general conclusions regarding the temperature dependence of the thermal conductivity. The present work investigates the relationship between the thermal conductivity of several nanouids and temperature over a wider temperature range than has been explored in previous studies. We have chosen to investigate nanouids based on ethylene glycol because it is known that the thermal conductivity versus temperature relationship for polar uids such as water and ethylene glycol is nonlinear and exhibits a maximum [14,15]. Therefore, one outcome of the present investigation is expected to be an elucidation of the role played by the base uid on the thermal conductivity of nanouids. In addition, nanouids containing three different mass fractions of alumina nanoparticles were investigated in order to ascertain whether dispersions of 20 nm size nanoparticles exhibit the anomalous behavior in thermal conductivity that has been reported in the literature. 2. Models for thermal conductivity enhancement Maxwell [1] derived the following relationship for the thermal conductivity of dilute dispersions of spherical particles: k= kp + 2k1 + 2(kp k1 ) k1 kp + 2k1 (kp k1 ) (1)

of the base uid, although size dependence is added via the parameter . Kumar et al. [10] attributed the thermal conductivity enhancement to Brownian motion and obtained a relationship that accounts for the effects of both particle size and temperature. Their model can be represented by the following equation: r1 k k1 = c up k1 k1 rp (1 ) (5)

where c is an adjustable parameter, r1 is the radius of the molecules of the base uid, rp the radius of the nanoparticles, and up the average particle velocity derived from the StokesEinstein equation as follows: up = 8k b T 2 rp (6)

where k is the thermal conductivity of the dispersion, kp the thermal conductivity of the particles, k1 the thermal conductivity of the pure uid, and is the volume fraction of particles. Hamilton and Crosser [2] extended this model to include an empirical factor n to account for the shape of the (nonspherical) particles: k= kp + (n 1)k1 + (n 1)(kp k1 ) k1 kp + (n 1)k1 (kp k1 ) (2)

In Eq. (6), kb is Boltzmanns constant, T is the temperature, and the viscosity of the uid. The Kumar et al. model imposes an additional temperature dependence to that of the base uid via up . Also, the thermal conductivity enhancement increases 3 with particle size as 1/rp In a more recent paper, Jang and Choi [11] described the thermal conductivity of nanouids in terms of three contributions attributed to the pure uid, to the particles, and to microconvection caused by motion of the particles as follows: k = k1 (1 ) + C1 kp + 3 C2 r1 k1 Rep2 Pr, rp (7)

where C1 is a constant related to the Kapitza resistance, C2 is another constant, Pr is the Prandtl number, and Rep the Reynolds number dened as: Rep = 1 kb T 32 l1 (8)

with n = 3 for spheres, and n = 6 for cylinders. Neither the Maxwell nor the Hamilton and Crosser models contain any dependence on particle size, and they imply that the temperature dependence of the thermal conductivity is approximately the same as that of the base uid. Yu and Choi [9] proposed an extension of the Maxwell equation that incorporates the effect of particle size via a contribution from an ordered liquid layer at the solidliquid interface. This ordered liquid layer is assumed to exhibit a greater thermal conductivity than the bulk liquid, but less than that of the solid. An effective thermal conductivity of particles is used in the Maxwell equation which becomes: k= kpe + 2k1 + 2(kpe k1 )(1 + )3 k1 kpe + 2k1 (kpe k1 )(1 + )3 (3)

In Eq. (8), 1 is the density of the liquid, and l1 the mean free path of a liquid molecule. Once again, an additional thermal conductivity contribution arises from the last term, as well as a dependence on particle size as 1/rp 3. Experimental Several nanouids (uids containing specied mass fractions of nanoparticles) were prepared by dispersing pre-weighed quantities of alumina particles in ethylene glycol at ambient conditions. The mixtures were subjected to ultrasonic mixing for several minutes to obtain uniform dispersions, and the resulting dispersions remained uniform for the duration of the experiments because of surface charges on the particles. The alumina nanoparticles were purchased from Nanostructured & Amorphous Materials, Inc., with a primary particle size of 20 nm reported by the vendor. A liquid metal transient hot wire device was used to measure the thermal conductivity of each nanouid. The apparatus and procedure have been described by Bleazard and Teja [1620] who have also reported measurements of the thermal conductivity of a variety of electrically conducting uids using this

where is the ratio of the ordered layer thickness to the nanoparticle radius, and kpe is the effective thermal conductivity of the particles dened as: kpe = 2(1 ) + (1 + )3 (1 + 2 ) (1 ) + (1 + )3 (1 + 2 ) kp (4)

where is the ratio of the thermal conductivity of the ordered layer to that of the solid particle. Once again, the temperature dependence of the model is approximately the same as that

M.P. Beck et al. / Fluid Phase Equilibria 260 (2007) 275278

277

apparatus at temperatures as high as 465 K. Briey, a mercurylled glass capillary is suspended in the uid or dispersion, with the glass capillary serving to insulate the mercury hot-wire from the electrically conducting uid or dispersion. The mercury wire forms one resistor in a Wheatstone bridge circuit and is heated when a constant voltage is applied to the bridge. The temperature rise of the wire is calculated from the change in the resistance of the mercury with time, obtained by measuring the voltage offset of the initially balanced Wheatstone bridge. The temperature rise versus time data are then used to calculate the thermal conductivity by solving Fouriers equation for an innite line heat source in an innite medium: T = q ln 4k 4t 2C rw (9)

Table 1 Thermal conductivity data for nanouids studied in this work Mass fraction of Al2 O3 (%) 3.26 3.26 3.26 3.26 3.26 3.26 9.34 9.34 9.34 9.34 9.34 9.34 12.2 12.2 12.2 12.2 12.2 12.2 T (K) 302.0 323.4 347.3 372.2 392.4 411.1 296.3 323.6 349.0 373.3 392.1 409.6 304.0 323.7 348.5 373.3 391.0 409.0 Volume fraction of Al2 O3 (%) 1.00 0.98 0.97 0.95 0.94 0.92 3.00 2.95 2.90 2.85 2.81 2.78 3.99 3.94 3.87 3.81 3.76 3.71 Mean k (W m1 K1 ) 0.258 0.259 0.262 0.267 0.264 0.260 0.276 0.282 0.284 0.285 0.287 0.280 0.290 0.291 0.294 0.293 0.288 0.285 Standard deviation 0.004 0.004 0.001 0.003 0.002 0.002 0.003 0.002 0.005 0.005 0.003 0.007 0.005 0.005 0.003 0.006 0.007 0.007

where T is the temperature rise of the wire, q the heat dissipation per unit length, k the thermal conductivity of the uid, t the time from the start of heating, the thermal diffusivity of the uid, rw the radius of the wire, and C the exponent of Eulers constant. A linear relationship between the temperature rise of the wire and the natural log of time is used to conrm that conduction is the primary mode of heat transfer during the measurement. Corrections to the temperature rise are made for the nite characteristics of the wire, the insulating layer around the wire, the nite extent of the uid, and heat loss due to radiation. Finally, an effective wire length is obtained by calibrating the instrument with a reference uid in order to account for end effects and the nonuniform thickness of the capillary. In the present work, the calibration was performed using IUPAC suggested values for the thermal conductivity of water [14] and dimethyl phthalate [21]. Additional details of the apparatus and technique are available elsewhere [20]. Each value of the thermal conductivity was obtained from an average of 5 measurements with an estimated accuracy of 2% and a reproducibility of 1%. The highest temperature measured in the present work was 411 K because of the onset of convection in the apparatus. The measurements were performed by placing the cell in a Techne constant temperature uidized sand bath (model SBL2D). 4. Results and discussion Our thermal conductivity measurements are presented in Table 1. Each data point represents an average of ve measurements at a specied mass fraction and temperature. Standard deviations for each data point are also given. Note that although the mass fraction remained constant for a sample over a range of temperatures, the volume fraction of the sample decreased with temperature because of the change in density of the base uid with temperature. The data are plotted in Fig. 1, together with the thermal conductivity data for pure ethylene glycol from the literature [15]. The results clearly demonstrate curvature in the thermal conductivity versus temperature behavior for both the base uid and each nanouid, with each nanouid displaying a maximum in thermal conductivity at approximately the same temperature as the maximum observed in the pure base uid. These results suggest that the thermal conductivity versus tem-

perature behavior of the nanouid closely follows the behavior of the base uid. The models of Kumar et al. [10] and Jang and Choi [11] attribute the thermal conductivity enhancement to Brownian motion resulting in greater temperature dependence than that due to the base uid. Furthermore, the enhancement increases monotonically with temperature. Our results clearly demonstrate that the thermal conductivity closely follows the temperature behavior of the base uid within the standard deviation of the data. This conclusion is supported by the analysis of Evans et al. [22] based on kinetic theory which also suggests that Brownian motion may play a minor role in determining the thermal conductivity enhancement of nanouids. Maxwell type models therefore are likely to be superior in describing the dependence of the thermal conductivity on temperature, since most of the temperature dependence in these models arises from the base uid.

Fig. 1. Thermal conductivity of pure ethylene glycol and 3 nanouids consisting of ethylene glycol and a specied mass fraction of 20 nm alumina nanoparticles. Dashed lines represent the HamiltonCrosser equation (with n = 3.4).

278

M.P. Beck et al. / Fluid Phase Equilibria 260 (2007) 275278

the same temperature as the base uid. This dependence cannot be reproduced by models that attribute the thermal conductivity enhancement in nanouids to Brownian motion, and is best represented by Maxwell type models. Our results also indicate that Maxwell type models such as that of Hamilton and Crosser or Yu and Choi with one adjustable parameter are able to successfully correlate the data, suggesting that the ethylene glycol based nanouids studied in this work do not exhibit anomalous thermal conductivity behavior. The value of the shape factor (n = 3.4) obtained by tting our data with the Hamilton and Crosser model suggests that some clustering of nanoparticles occurs during our experiments and the resulting clusters are not spherical. References
[1] J.C. Maxwell, A Treatise on Electricity and Magnetism, third ed., Oxford University Press, London, 1892. [2] R.L. Hamilton, O.K. Crosser, Ind. Eng. Chem. Fundam. 1 (1962) 187. [3] X. Wang, X. Xu, S.U.S. Choi, J. Thermophys. Heat Transfer. 13 (1999) 474480. [4] J.A. Eastman, S.U.S. Choi, S. Li, W. Yu, L.J. Thompson, Appl. Phys. Lett. 78 (2001) 718720. [5] S.U.S. Choi, Z.G. Zhang, W. Yu, F.E. Lockwood, E.A. Grulke, Appl. Phys. Lett. 79 (2001) 22522254. [6] M.S. Liu, M.C.C. Lin, I.T. Huang, C.C. Wang, Chem. Eng. Technol. 29 (2006) 7277. [7] H. Xie, J. Wang, T. Xi, Y. Liu, F. Ai, J. Mater. Sci. Lett. 21 (2002) 193 195. [8] H. Xie, J. Wang, T. Xi, Y. Liu, F. Ai, Q. Wu, J. Appl. Phys. 91 (2002) 45684572. [9] W. Yu, S.U.S. Choi, J. Nanopart. Res. 5 (2003) 167171. [10] D.H. Kumar, H.E. Patel, V.R.R. Kumar, T. Sundararajan, T. Pradeep, S.K. Das, Phys. Rev. Lett. 93 (2004) 144301. [11] S.P. Jang, S.U.S. Choi, Appl. Phys. Lett. 84 (2004) 43164318. [12] S.K. Das, N. Putra, P. Thiesen, W. Roetzel, J. Heat Transfer. 125 (2003) 567574. [13] X. Zhang, H. Gu, M. Fujii, J. Appl. Phys. 100 (2006) 044325. [14] C.A. Meyer, ASME Steam Tables: Thermodynamics and Transport Properties of Steam, sixth ed., American Society of Mechanical Engineering, New York, 1993. [15] R. Diguilio, A.S. Teja, J. Chem. Eng. Data 35 (1990) 117121. [16] T.F. Sun, A.S. Teja, J. Chem. Eng. Data 49 (2004) 13111317. [17] T. Sun, A.S. Teja, J. Chem. Eng. Data 49 (2004) 18431846. [18] T.F. Sun, A.S. Teja, J. Chem. Eng. Data 48 (2003) 198202. [19] J.G. Bleazard, T.F. Sun, A.S. Teja, Int. J. Thermophys. 17 (1996) 111125. [20] J.G. Bleazard, A.S. Teja, J. Chem. Eng. Data 40 (1995) 732737. [21] K.N. Marsh, Recommended Reference Materials for the Realization of Physicochemical Properties, Blackwell Scientic Publications, Boston, 1987. [22] W. Evans, J. Fish, P. Keblinski, Appl. Phys. Lett. 88 (2006) 093116.

Fig. 2. Thermal conductivity at 373 K of nanouids consisting of ethylene glycol and 20 nm alumina nanoparticles. Circles represent experimental data of this work. The solid line represents calculations using the Maxwell equation whereas the dashed line represents calculations using the Hamilton and Crosser equation with n = 3.4.

Fig. 2 demonstrates our results showing the dependence of volume fraction of particles on the thermal conductivity. The behavior is linear and can be reproduced using the Hamilton and Crosser model with n = 3.4. This value of the shape factor was obtained from a best t of experimental data at all temperatures, and suggests that agglomeration of particles occurs leading to a mean shape that is nonspherical, since the value of n lies between that of spherical and cylindrical particles. The data may also be tted with the Yu and Choi model, if the thermal conductivity of the ordered liquid layer is assumed to be the same as that of the particles and the thickness of the layer is set equal to 0.4 nm by linear least squares to provide the best t for all data. However, the calculations are very sensitive to the thickness of the ordered liquid layer and its thermal conductivity. 5. Conclusions A liquid metal transient hot wire apparatus was used to measure the thermal conductivity of ethylene glycol based alumina nanouids at temperatures ranging from 298 to 411 K. These data span the widest range of temperatures hitherto measured for any nanouid and show that ethylene glycol based nanouids exhibit a maximum in thermal conductivity at approximately

You might also like