You are on page 1of 3

APPLIED PHYSICS LETTERS VOLUME 78, NUMBER 6 5 FEBRUARY 2001

Anomalously increased effective thermal conductivities of ethylene


glycol-based nanofluids containing copper nanoparticles
J. A. Eastmana)
Materials Science Division, Argonne National Laboratory, Argonne, Illinois 60439
S. U. S. Choi
Energy Technology Division, Argonne National Laboratory, Argonne, Illinois 60439
S. Lib)
Materials Science Division, Argonne National Laboratory, Argonne, Illinois 60439
W. Yu
Energy Technology Division, Argonne National Laboratory, Argonne, Illinois 60439
L. J. Thompson
Materials Science Division, Argonne National Laboratory, Argonne, Illinois 60439
共Received 21 September 2000; accepted for publication 21 November 2000兲
It is shown that a ‘‘nanofluid’’ consisting of copper nanometer-sized particles dispersed in ethylene
glycol has a much higher effective thermal conductivity than either pure ethylene glycol or ethylene
glycol containing the same volume fraction of dispersed oxide nanoparticles. The effective thermal
conductivity of ethylene glycol is shown to be increased by up to 40% for a nanofluid consisting of
ethylene glycol containing approximately 0.3 vol % Cu nanoparticles of mean diameter ⬍10 nm.
The results are anomalous based on previous theoretical calculations that had predicted a strong
effect of particle shape on effective nanofluid thermal conductivity, but no effect of either particle
size or particle thermal conductivity. © 2001 American Institute of Physics.
关DOI: 10.1063/1.1341218兴

Heating or cooling fluids are of major importance to is desirable to use particles with a large surface area-to-
many industrial sectors, including transportation, energy sup- volume ratio. Hamilton and Crosser2 focused on the possible
ply and production, and electronics. The thermal conductiv- effects of increasing particle surface area by controlling par-
ity of these fluids plays a vital role in the development of ticle shapes to be nonspherical and modified Maxwell’s
energy-efficient heat transfer equipment. However, conven- model for spherical particles. However, typically less than an
tional heat transfer fluids have poor heat transfer properties order-of-magnitude improvement in surface area per particle
compared to most solids. Despite considerable previous re- volume is attainable experimentally by this strategy alone.
search and development focusing on industrial heat transfer On the basis of this historical background, it was pro-
requirements, major improvements in heat transfer capabili- posed that if nanometer-sized particles could be suspended in
ties have been lacking. As a result, a clear need exists to traditional heat transfer fluids, a new class of engineered flu-
develop new strategies for improving the effective heat trans- ids with high thermal conductivity could be produced.9
fer behavior of conventional heat transfer fluids. These so-called ‘‘nanofluids’’ are expected to exhibit supe-
Crystalline solids have thermal conductivities that are rior properties relative to those not only of conventional heat
typically larger than those of fluids by 1–3 orders of magni- transfer fluids, but also of fluids containing micrometer-sized
tude. Therefore, fluids containing suspended solid particles metallic particles. Since the surface-area-to-volume ratio is
can reasonably be expected to display significantly enhanced 1000 times larger for particles with a 10 nm diameter than
thermal conductivities relative to those of pure fluids. Nu- for particles with a 10 ␮m diameter, a much more dramatic
merous theoretical and experimental studies of the effective improvement in effective thermal conductivity is expected as
thermal conductivity of liquids containing suspended solid a result of decreasing the particle size in a suspension than
particles have been conducted since Maxwell’s theoretical can be obtained by altering the shapes of larger particles.
work1 was first published more than 100 years ago.2–8 How- Recently, we demonstrated that nanofluids consisting of
ever, with very few exceptions, previous studies of the ther- CuO or Al2O3 nanoparticles in water or ethylene glycol ex-
mal conductivity of suspensions have been confined to those hibit enhanced thermal conductivity.10 A maximum increase
containing millimeter- or micrometer-sized particles. in thermal conductivity of approximately 20% was observed
Maxwell’s model1 predicts that the effective thermal in that study for 4 vol % CuO nanoparticles with average
conductivity of suspensions containing spherical particles in- diameter 35 nm dispersed in ethylene glycol. Similar behav-
creases with the volume fraction of the solid particles. Be- ior was observed in another recent study of Al2O3 nanopar-
cause heat transfer takes place at the surface of the particle, it ticles dispersed in water by Masuda and co-workers.11 The
present work demonstrates that significantly larger improve-
a兲
Electronic mail: jeastman@anl.gov ments in effective thermal conductivity are obtained for
b兲
Present address: Seagate Technology, Bloomington, MN. nanofluids containing smaller sized and higher conductivity

0003-6951/2001/78(6)/718/3/$18.00 718 © 2001 American Institute of Physics


Downloaded 04 Jul 2006 to 202.63.233.8. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
Appl. Phys. Lett., Vol. 78, No. 6, 5 February 2001 Eastman et al. 719

FIG. 1. Bright-field transmission electron micrograph of Cu nanoparticles


produced by direct evaporation into ethylene glycol. Very little agglomera-
tion occurs using this processing method.

copper nanoparticles. As will be described, this is a surpris-


ing result based on theoretical predictions. FIG. 2. The effective thermal conductivity of ethylene glycol is seen to be
A one-step procedure for producing nanofluids contain- improved by up to 40% through the dispersion of 0.3 vol % Cu nanopar-
ing metallic particles was used to disperse nanocrystalline ticles. Linear fits to the data are shown as a guide to the eye. The largest
increase in conductivity was seen for a nanofluid that also contained a small
copper particles into ethylene glycol with little agglomera-
quantity of thioglycolic acid to improve the stability of the metal particles
tion. Briefly, this technique, invented by Akoh and against settling. A small effect of sample age on thermal conductivity was
co-workers,12 involves the direct condensation of metallic also observed. Samples without thioglycolic acid tested within two days of
vapor into nanoparticles by contact with a flowing low vapor preparation are denoted ‘‘fresh,’’ while those stored up to two months prior
to testing are denoted ‘‘old.’’
pressure liquid. A modification of the direct-condensation
process developed by Wagener et al.13 was used in the cur- ids as a function of nanoparticle volume fraction. Data are
rent experiments 共see Ref. 14 for further details兲. As seen in plotted normalized to the conductivity of nonparticle-
Fig. 1, Cu nanoparticles with little agglomeration and an containing ethylene glycol. Several important points can be
average diameter of less than 10 nm were produced. Load- noted. First, very significant increases in thermal conductiv-
ings of up to approximately 0.5 vol % were produced and ity are seen for all measured nanofluids, with conductivity
tested. Particle loadings were estimated by weighing the re- enhancements of up to 40% observed for particle loadings
sistive evaporation source before and after the nanofluid well below one volume per cent. Second, nanofluids contain-
preparation. For some samples, a small amount of thiogly- ing thioglycolic acid as a stabilizing agent show improved
colic acid 共⬍1 vol %兲 was added to the nanofluid to improve behavior compared to nonacid-containing nanofluids. It
the particle dispersion behavior. should be noted that fluids containing thioglycolic acid, but
The transient hot-wire 共THW兲 method15–17 was used in no particles, showed no improvement in thermal conductiv-
this study to measure fluid thermal conductivity. A THW ity. Third, fresh nanofluids tested within two days of prepa-
system involves a wire 共typically platinum兲 suspended sym- ration exhibited slightly higher conductivities than fluids that
metrically in a liquid in a vertical cylindrical container. Na- were stored up to two months prior to measurement.
gasaka and Nagashima’s method,18 in which the wire is Figure 3 shows comparisons between metallic and oxide
coated with a thin electrical insulation layer, was used in the nanofluids. All nanofluids containing Cu nanoparticles di-
current experiment to avoid problems associated with the rectly dispersed in ethylene glycol show significantly larger
measurement of electrically conducting fluids. Briefly, the increases in thermal conductivity than nanofluids containing
THW technique works by measuring the temperature/time oxide particles produced by a two-step process involving
response of the wire to an abrupt electrical pulse. The wire is gas-condensation processing of powders followed by disper-
used as both heater and thermometer and the thermal con- sion into the fluid.10,11 The conductivity enhancements for
ductivity, k is calculated from a derivation of Fourier’s Law both metallic and oxide nanofluids are approximately linear

k⫽
q
4 ␲ 共 T 2 ⫺T 1 兲 冉冊
ln
t2
t1
, 共1兲
with particle volume %. Compared to previous studies of
oxide-containing nanofluids there are two potentially impor-
tant differences: 共1兲 the Cu particles used in the current study
where q is the applied electric power and T 1 and T 2 are the are expected to have a much higher intrinsic thermal conduc-
temperatures at times t 1 and t 2 . From the temperature coef- tivity than the oxide particles used previously, and 共2兲 the
ficient of the wire’s resistance, the temperature rise of the average particle diameters in the present study are more than
wire can be determined by the change in its electrical resis- four times smaller than for the oxide particles.
tance with time. Calibration experiments were performed for Hamilton and Crosser’s analysis predicts that the con-
ethylene glycol in the temperature range of 290–310 K and ductivity of two-component mixtures can be described by

冋 册
at atmospheric pressure. Literature values19 were reproduced
with an error of ⬍1.5%. k m ⫹ 共 n⫺1 兲 k 0 ⫺ 共 n⫺1 兲 ␣ 共 k 0 ⫺k m 兲
k⫽k 0 , 共2兲
Figure 2 shows the thermal conductivity of Cu nanoflu- k m ⫹ 共 n⫺1 兲 k 0 ⫹ ␣ 共 k 0 ⫺k m 兲
Downloaded 04 Jul 2006 to 202.63.233.8. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
720 Appl. Phys. Lett., Vol. 78, No. 6, 5 February 2001 Eastman et al.

even larger for metallic nanofluids than oxide nanofluids.


This suggests that current models, which account only for
the volume fraction and shape of the suspended particles and
the differences between the thermal conductivity of particles
and fluids, are insufficient to explain the energy transfer pro-
cesses in the nanofluid system. A more comprehensive
theory is needed to explain the behavior of nanofluids. Ad-
ditional studies are required before this theory can be devel-
oped and the mechanism共s兲 responsible for the observed en-
hancements in fluid thermal conductivity are understood.
These include planned experimental studies of samples with
varying particle sizes, but the same composition. Addition-
ally, more sophisticated theoretical treatments are needed.
Simulation studies have recently been initiated20 that hope-
fully will provide insight into the atomistic processes con-
trolling the thermal conductivity of nanofluids.
In summary, nanofluids consisting of Cu nanoparticles
directly dispersed in ethylene glycol have been observed to
exhibit significantly improved thermal conductivity enhance-
FIG. 3. Significantly greater enhancements are seen for nanofluids consist-
ing of ⬍10 nm diameter Cu nanoparticles dispersed into ethylene glycol ments compared to nonparticle-containing fluids or nanoflu-
than for ethylene glycol-based nanofluids containing either CuO or Al2O3 ids containing oxide particles. The large improvement in ef-
nanoparticles of average diameter 35 nm. This is an unexpected result since fective conductivity obtained for nanofluids containing
theoretical calculations had predicted no effect of either particle diameter or metallic particles holds significant potential for revolutioniz-
particle conductivity on nanofluid conductivity.
ing industries that are dependent on the performance of heat
transfer fluids.
where k is the mixture thermal conductivity, k 0 is the liquid
thermal conductivity, k m is the thermal conductivity of solid This work was supported by the U.S. Department of En-
particles, ␣ is the particle volume fraction, and n is an em- ergy, Office of Science and Office of Transportation Tech-
pirical scaling factor that takes into account the effect on nologies, under Contract No. W-31-109-Eng-38.
thermal conductivity of different particle shapes.
Since the nanoparticles produced in this investigation are
approximately spherical, n⬇3 and Eq. 共2兲 thus predicts very 1
J. C. Maxwell, A Treatise on Electricity and Magnetism, 2nd ed. 共Oxford
low enhancements in thermal conductivity compared to those University Press, Cambridge, 1904兲, pp. 435–441.
2
R. L. Hamilton and O. K. Crosser, I & EC Fundamentals 1, 187 共1962兲.
observed 共e.g., a conductivity ratio of approximately 1.015 3
Z. Hashin and S. Shtrikman, J. Appl. Phys. 33, 3125 共1962兲.
would be predicted for a 0.5 vol % copper particles in ethyl- 4
D. J. Jeffrey, Proc. Phys. Soc., London, Sect. A 335, 355 共1973兲.
ene glycol, which is a smaller enhancement than the ob- 5
D. J. Jackson, Classical Electrodynamics, 2nd ed. 共Wiley, London, 1975兲.
served value by well over an order of magnitude兲. An obvi-
6
R. H. Davis, Int. J. Theor. Phys. 7, 609 共1986兲.
7
R. R. Bonnecaze and J. F. Brady, Proc. Phys. Soc., London, Sect. A 432,
ous shortcoming in Hamilton and Crosser’s theory is the lack 445 共1991兲.
of any predicted dependence of conductivity on particle size. 8
S. Lu and H. Lin, J. Appl. Phys. 79, 6761 共1996兲.
9
While the shape factor in the Hamilton and Crosser analysis U. S. Choi, in Developments and Applications of Non-Newtonian Flows,
takes into account the increased surface area of nonspherical edited by D. A. Siginer and H. P. Wang 共The ASME, New York, 1995兲,
Vol. 231/MD-Vol. 66, pp. 99–105.
particles, it does not account for the orders-of-magnitude in- 10
S. Lee, U. S. Choi, S. Li, and J. A. Eastman, ASME J. Heat Transfer 121,
crease in surface-area-to-volume ratio that accompany de- 280 共1999兲.
11
creasing particle size into the nanocrystalline regime. H. Masuda, A. Ebata, K. Teramae, and N. Hishinuma, Netsu Bussei 4, 227
A second possible weakness in the Hamilton–Crosser 共1993兲.
12
H. Akoh, Y. Tsukasaki, S. Yatsuya, and A. Tasaki, J. Cryst. Growth 45,
analysis is that the thermal conductivity of the particles has 495 共1978兲.
only a weak predicted effect on k in Eq. 共2兲. For example, 13
M. Wagener, B. S. Murty, and B. Günther, in Nanocrystalline and Nano-
using quantities in Eq. 共2兲 that correspond to 0.5 vol % par- composite Materials II, edited by S. Komarnenl, J. C. Parker, and H. J.
ticles in ethylene glycol predicts less than 0.1% larger con- Wollenberger 共Materials Research Society, Pittsburgh PA, 1997兲, Vol.
457, pp. 149–154.
ductivity ratio for Cu particles than for Al2O3 particles. Since 14
J. A. Eastman, U. S. Choi, S. Li, L. J. Thompson, and S. Lee, in Nano-
in our previous study10 nanofluids containing essentially crystalline and Nanocomposite Materials II, edited by S. Komarnenl, J. C.
identically sized and shaped oxide particles of different com- Parker, and H. J. Wollenberger 共Materials Research Society, Pittsburgh
PA, 1997兲, Vol. 457, pp. 3–11.
position were observed to exhibit conductivity ratios that 15
J. Kestin and W. A. Wakeham, Physica A 92, 102 共1978兲.
varied by almost 100%, this weak predicted effect of k m on k 16
H. M. Roder, J. Res. Natl. Bur. Stand. 86, 457 共1981兲.
appears unreasonable. 17
A. I. Johns, A. C. Scott, J. T. R. Watson, and D. Ferguson, Philos. Trans.
It was demonstrated that thermal conductivities pre- R. Soc. London, Ser. A 325, 295 共1988兲.
18
Y. Nagasaka and A. Nagashima, J. Phys. E 14, 1435 共1981兲.
dicted by theoretical models such as Hamilton and Crosser’s 19
In Thermal Properties of Matter. The TPRC Data Series, edited by Y. S.
are much lower than the measured data for oxide Touloukian, and C. Y. Ho 共Plenum, New York, 1970–1977兲.
nanofluids.18 The present study shows the discrepancy is 20
P. Keblinski and S. R. Phillpot 共private communication兲.

Downloaded 04 Jul 2006 to 202.63.233.8. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp

You might also like