REPORTS

16. D. Y. Chung et al., Science 287, 1024 (2000).
17. Materials and methods are available as supporting
material on Science Online.
18. D. Y. Chung et al., Chem. Mater. 9, 3060 (1997).
19. These data were obtained with the flash diffusivity–
specific heat method at the Thermophysical Proper-
ties Research Laboratory, West Lafayette, IN (www.
tprl.com/). The sample configuration was in the form
of a flat disk of ⬃14-mm diameter and 2.5-mm
thickness.
20. Hall measurements of samples with m ⫽ 18 and
1 ⫺ x ⫽ 0.55 and 0.67 were carried out as a function
of temperature. The carrier concentrations (elec-
trons) at room temperature were ⬃2.15 ⫻ 1019 and
⬃1.9 ⫻ 1019 cm⫺3, respectively. Carrier concentra-
tions rise sharply with falling temperature, suggesting
a narrowing of the band gap in these materials. This
is opposite to what occurs in most semiconductors
(in which the band gap widens with falling temper-
ature) and similar to what occurs in PbTe (23). At the
same time, the electron mobilities are markedly high
for both x ⫽ 0.55 and 0.67 samples at ⬃800 cm2/V䡠s
and similar to those reported for the MBE-grown
thin-film superlattice PbSe/PbTe systems (15).
21. D. Bilc et al., in preparation.
22. The power conversion efficiency of a thermo-
element is given by
␩ ⫽ (⌬T/T hot)[(1⫹ZT ave)1/2–1]/
[(1⫹ZT ave)1/2⫹(T cold /T hot)]
where ⌬T is the temperature difference across the
device. The first fraction in this expression is the
Carnot efficiency. ZT is defined in (1).
23. Y. I. Ravich, B. A. Efimova, I. A. Smirnov, Semiconduct-
ing Lead Chalcogenides, T. S. Stil’bans, Ed. (Plenum,
New York, 1970), pp. 323–346.
24. L. Genzel, Z. Phys. 135, 177 (1953).
25. S. Loo, T. Hogan, in preparation.
26. Financial support from the Office of Naval Research
and Defense Advanced Research Projects Agency is
gratefully acknowledged.
Supporting Online Material
www.sciencemag.org/cgi/content/full/303/5659/818/
DC1
Materials and Methods
Fig. S1
27 October 2003; accepted 11 December 2003

Superlattices of Iron Nanocubes be controlled. The latter point is of special

interest for further assembling the particles

Synthesized from Fe[N(SiMe3)2]2 into 2D or 3D superlattices. After exploring

several organometallic complexes, we con-
sidered amido complexes because such
Frédéric Dumestre,1,2 Bruno Chaudret,1* Catherine Amiens,1 compounds have previously been shown to
Philippe Renaud,2 Peter Fejes3 be versatile precursors for the production of
particles of industrial interest (13) or crys-
The reaction of the metal-organic precursor Fe[N(SiMe3)2]2 with H2 in the talline 3D superlattices (14 ). We now find
presence of a long-chain acid and a long-chain amine in various proportions that the derivative Fe[N(SiMe3)2]2 (where
produces monodisperse zerovalent iron nanoparticles. These Fe particles Me ⬅ -CH3) reported by Andersen (15)
display magnetic properties that match those of bulk iron as evidenced by fulfills all requirements.
magnetic and Mössbauer measurements. The nanoparticles adopt a cubic The decomposition of Fe[N(SiMe3)2]2
shape with edges of 7 nanometers and are incorporated into extended was carried out in solution at 150°C under
crystalline superlattices containing nanocubes in close proximity and with a dihydrogen atmosphere in the presence of
their crystallographic axes aligned. These superlattices are formed in so- hexadecylamine (HDA) and a long-chain
lution, precipitate in high yield, and may be redissolved and redeposited as acid (oleic acid, OA, or hexadecylamonium
two-dimensional arrays. chloride, HDAC). In a standard reaction, 1
mmol of Fe[N(SiMe3)2]2 was dissolved in
The assembly of nanoparticles in electronic de- cles displaying magnetic properties similar to 20 mL of mesitylene and reacted at 150°C
vices (1) appears as a bottom-up alternative to gas phase clusters in ultrahigh vacuum (5), with for 48 hours with 3 bar of H2 (initial pres-
present lithography techniques for the fabrica- control over their shape and hence magnetic sure at room temperature in the reactor) in
tion of ever smaller devices in microelectronics anisotropy (6, 7), and that they may be assem- the presence of 2 equivalents (eq) of HDA
and for the magnetic-storage industry. For ex- bled into 2D or 3D superlattices (8). However, and 1 eq acid. In both cases, a black pre-
ample, magnetic hard drives presently use sen- this does not apply to iron or to iron-based cipitate forms in ⬃50% yield, which is
sitive read heads formed of thick multilayers of alloys. Chemical syntheses of iron nanopar- shown by transmission electron microscopy
metal nanoparticles (2). However, new applica- ticles have used Fe(CO)5 as a precursor. For (TEM) and scanning electron microscopy
tions such as single electron devices (3) and example, upon sonolysis of Fe(CO)5 in the (SEM) to consist of iron nanocubes includ-
spin– dependent tunneling (4) require a fine- presence of polyvinylpyrrolidone or oleic acid ed in extended 3D superlattices (Fig. 1A,
tuning of the physical properties of the nano- (9), amorphous iron nanoparticles are produced Fig. 2, and fig. S1). The superlattices dis-
particles and, in turn, a nanometer-scale control in good yields. Similarly, Hyeon has described play defined shapes (cubes, parallelograms)
of the material, in terms of size, shape, compo- the synthesis of elongated bcc iron particles with extended faces in the micron range as
sition, and crystal structure, as well as an orga- (10) but, in most cases, the magnetization was shown by SEM (Fig. 2B and fig. S1). The
nization into two-dimensional (2D) or 3D either found depleted (9, 11) or not reported nanocubes adopt the bcc structure of bulk
nanocrystal superlattices. (10). A recent paper, however, describes a high iron as evidenced by x-ray diffraction (fig.
These devices also require high perfor- temperature decomposition of Fe(CO)5 in the S2) and by a selected area electron diffrac-
mance of individual nanoparticles, namely high presence of oleic acid and/or oleylamine lead- tion experiment on one of these superlat-
magnetization and adjustable anisotropy. Re- ing to nanoparticles of controlled size and mag- tices; discrete sharp spots are observed that
cent results have shown that chemical synthetic netization up to 200 A m2 kgFe⫺1 (12). confirm both the bcc structure of the
methods, in particular those based on the use of Iron-based materials are highly desir- nanocubes and the alignment of all their
organometallic precursors, may provide parti- able for magnetic applications because of crystallographic axes (Fig. 1B). In the case
their high magnetization (FeCo, FeNi) and of the reaction in the presence of OA
1
Laboratoire de Chimie de Coordination du CNRS, wide range of magnetic anisotropy (FeNi, (HDA:OA; 2 :1, 1), the cube edges display
205, route de Narbonne, 31077 Toulouse Cédex 04, FePt). Thus, it was of interest to find an a mean size of 7 nm (␴ ⫽ 0.4) and an
France. 2Digital DNA Labs, Semiconductor Products easily accessible precursor that would de- interparticle spacing estimated at ⬃1.6 nm,
Sector, Motorola, le Mirail B.P. 1029, 31023 Toulouse
Cedex, France. 3Digital DNA Labs, Semiconductor
compose under mild conditions, leave res- whereas with HDAC (HDA:HDAC; 2 :1,
Products Sector, Motorola, 2100 East Elliot Road, idues that would not perturb the magnetic 2), the mean size of the cube edges is
Tempe, AZ 85824, USA. properties of the resulting particles, and measured at 8.3 nm (␴ ⫽ 0.8), with an
*To whom correspondence should be addressed: therefore lead to nanoparticles, the size, interparticle spacing of ⬃2 nm. High-
chaudret@lcc-toulouse.fr shape, and surface ligands of which could resolution electron micrographs (HREM)

www.sciencemag.org SCIENCE VOL 303 6 FEBRUARY 2004 821

 REPORTS
of an assembly of nanocubes show that the
particles are not connected, confirm their
bcc structure, and evidence their air sensi-
tivity (fig. S3). If the reaction is carried out
at lower temperature or if the reaction is
stopped after 3 hours, a mixture of nano-
particles is obtained. These particles are
polymorph (spherical and cubic) and poly-
disperse in size. For an intermediate reac-
tion time (18 hours), cubic nanoparticles of
⬃6.2 nm are observed on the TEM grid.
These nanoparticles tend to organize into
2D superstructures (fig. S4).
Variations in the nature (use of dode-
cylamine, DDA, or palmitic acid, PA) and
relative concentration of amine and acid li-
gands (2 :1, 1:1, and 1: 2) have a limited
effect on the course of the reaction. Similar
superlattices of ⬃7 nm nanocubes were ob-
tained but in most cases were associated with
the presence of spherical nanoparticles of
smaller mean size (5 nm for 2 DDA: 1 PA).
Mössbauer spectra of the samples record-
ed at 80 K show a single sextet indicative of
the presence of a single species that displays
an isomeric shift of ␦ ⫽ 0.1092 mm s⫺1 and
a hyperfine field of 339 kOe attributed to
unoxidized Fe0 nanoparticles in a blocked
state (Fig. 3A). In addition, quasistatic mea-
Fig. 1. 3D superlattice of iron nanocubes as
deposited on a carbon-coated copper grid from
a toluene dispersion. (A) TEM micrograph of a
3D superlattice. Bar, 10 nm. (B) Diffraction
pattern from the superlattice presented in (A).
Fig. 3. Mössbauer and magnetic measurements. (A) Mössbauer spectrum registered at 80 K of iron
nanocubes in product 2. (B) Filled triangles, magnetization loop at 2 K of product 2. The solid line
represents the value of saturation magnetization for bulk iron (223 A m2 kgFe⫺1). (C) Inset showing
hysteretic behavior.
822 6 FEBRUARY 2004 VOL 303 SCIENCE www.sciencemag.org
surements of the magnetizations in a super-
conducting quantum interference device
(SQUID) magnetometer according to a zero
field cooling–field cooling procedure under a
low applied magnetic field (1 mT) show
blocking temperatures close to or slightly
higher than room temperature for both sam-
ples. These blocking temperatures are much
higher than those expected given the volume
of these cubes (⬃50 and 80 K for 1 and 2,
respectively). This probably results from both
shape anisotropy and dipolar antiferromag-
netic couplings as revealed by the low rema-
nent magnetization (83 and 28 A m2 kgFe⫺1
for 1 and 2, respectively). At 2 K, the mag-
netizations at 5T deduced from the hysteresis
curves were found close to bulk values, re-
spectively Ms ⫽ 212 A m2 kgFe⫺1 and 207 A
m2 kgFe⫺1 for 1 and 2 (bulk Fe, 223 A m2
kgFe⫺1). Coercive fields of 77 mT and 22 mT
were measured at 2 K, whereas almost no
coercivity was observed for both samples at
room temperature in agreement with the ap-
proach to the superparamagnetic limit at the
SQUID time scale (Fig. 3B).
Decomposition of Fe[N(SiMe3)2]2 in
such conditions is thus demonstrated to
lead to pure iron nanoparticles organized in
3D lattices. The very short interparticle
distance found in the superlattices (⬃1.6 to
2 nm) when compared with the ligand chain
length (⬃2.2 nm) rules out a simple self-
organization of ligand-protected particles
Fig. 2. Assembly of iron nanocubes. (A) TEM
micrograph of a 3D superlattice as observed
from thin slices of the product embedded in an
epoxy resin. Bar, 10 nm. (B) SEM micrograph of
a 3D superlattice of nanocubes. Bar, 500 nm.
on a microscopy grid. It is therefore con-
sistent with the crystallization of these cu-
bic arrays in solution in a way similar to
what has been observed for PbSe (16 ) and
Sn particles (14 ). Upon observation of
these superlattices through classical TEM,
it is not possible to find definitive evidence
for the shape of the individual particles,
which may be cubes or platelets as previ-
ously observed in the case of nanodisks of
cobalt that first appeared as nanorods (17 ).
The particle shape was therefore investigat-
ed through ultramicrotomy. In this case, the
particles are randomly oriented in the bind-
er and display images only of squares and
not of thin rectangles, which confirms their
cubic nature (Fig. 2). In addition, this ex-
periment shows that all particles are includ-
ed in 3D superlattices. These superlattices
may be redissolved in a tetrahydrofurane/
stearic acid mixture, allowing their depo-
sition as monolayers (fig. S5). The
nanocubes are highly air sensitive, which
leads to the formation of a crystalline mag-
netite shell upon exposure to air for a few
minutes, as evidenced by HREM (fig. S3).
We have shown that similar reaction
conditions promote the synthesis of cobalt
nanorods of homogeneous diameter and
length from the organometallic precursor
Co(␩3C8H13)(␩4C8H12) (6 ). We found ini-
tial formation of spherical nanoparticles, which
then would coalesce into nanorods. In the
present case, a similar process is suggested
from the observation of nanoparticles at the
early stage of the reaction. The growth of these
particles, possibly through coalescence, produc-
es nanocubes that initially are separated and
organized in two dimensions (fig. S4) and that
further spontaneously organize into 3D super-
structures as observed after 48 hours of reaction
time. The hollow matrix surrounding the parti-
cles at this reaction stage can be attributed
either to the acid/amine ligands or to interme-
diate iron carboxylate complexes that form dur-
ing the course of the reaction and decompose
very slowly, thus providing a constant source of
iron atoms in the medium (18). This result
suggests that the crystal structure of the metal
(bcc for Fe, hcp for Co) might participate in the
shape control of the mesoscopic nano-objects
under thermodynamic conditions. One way this
structural effect could be involved is through
ligand coordination on specific faces of the
growing nano-objects (17, 19). This process
would be isotropic for bcc iron and anisotropic
for hcp cobalt. It is noteworthy that nanopar-
ticles of Fe, Mo, and PbSe may also adopt a
cubic shape whether prepared by sputtering (20,
21) or in solution (22), which agrees with a
structure control on the overall shape of the
particles. In the case of cobalt nanoparticles, the
changes of amine and/or acid ligands were
shown to display a strong influence on the
length and diameter of the rods and hence on

their aspect ratio and magnetic properties. This
is not the case for iron because, whatever the
ligands used, nanocubes of ⬃7 nm form. Fur-
thermore, the interparticle distance in these
samples is uniform and very short (1.6 nm),
which tends to suggest the absence of coordi-
nated long-chain ligands at the surface of the
particles. This can result from the formation in
all cases of hexamethyldisilazane, which can
remain coordinated at the surface of the parti-
cles. It is also noteworthy that neither water
formed upon condensation of OA with HDA in
the reaction conditions nor chloride ions present
in HDAC are detrimental to the formation and
magnetic properties of the iron nanocubes.
References and Notes
1. S. Sun, C. B. Murray, D. Weller, L. Folks, A. Moser,
Science 287, 1989 (2000).
2. J. C. Mallinson, Magneto-Resistive Heads: Fundamental and
Applications (Academic Press, New York, 1996).
3. R. P. Andres et al., Science 272, 1323 (1996).
4. C. T. Black, C. B. Murray, R. L. Sandstrom, S. Sun,
Science 290, 1131 (2000).
5. M. Respaud et al., Phys. Rev. B 57, 2925 (1998).
6. F. Dumestre et al., Angew. Chem. Int. Ed. Engl. 41,
4286 (2002).
7. N. Cordente et al., Nano Lett. 1, 565 (2001).
8. F. Dumestre et al., Angew. Chem. Int. Ed. Engl. 42,
5213 (2003).
9. K. S. Suslick, M. Fang, T. Hyeon, J. Am. Chem. Soc.
118, 11960 (1996).
10. S. J. Park et al., J. Am. Chem. Soc. 122, 8581 (2000).
11. M. W. Grinstaff, M. B. Salamon, K. S. Suslick, Phys.
Rev. B 48, 269 (1993).
12. D. Farrell, S. A. Majetich, J. P. Wilcoxon, J. Phys. Chem.
B 107, 11022 (2003).
13. C. Nayral et al., Chem. Eur. J. 6, 4082 (2000).
14. K. Soulantica, A. Maisonnat, M. C. Fromen, M. J.
Casanove, B. Chaudret, Angew. Chem. Int. Ed. Engl.
42, 1945 (2003).
15. R. A. Andersen et al., Inorg. Chem. 27, 1782 (1988).
16. C. B. Murray et al., IBM J. Res. Dev. 45, 47 (2001).
17. V. F. Puntes, D. Zanchet, C. K. Erdonmez, A. P. Alivi-
satos, J. Am. Chem. Soc. 124, 12874 (2002).
18. Z. A. Peng, X. Peng, J. Am. Chem. Soc. 123, 1389
(2001).
REPORTS
19. V. F. Puntes, K. M. Krishnan, A. P. Alivisatos, Science
291, 2115 (2001).
20. H. K. Lee et al., J. Appl. Phys. 93, 7047 (2003).
21. A. S. Edelstein, G. M. Chow, E. I. Altman, R. J. Colton,
D. M. Hwang, Science 251, 1590 (1991).
22. C. B. Murray, K. L. Stokes, K.-S. Cho, Abstracts of
Papers, 225th American Chemical Society National
Meeting, New Orleans, LA, 23 to 27 March 2003
(2003).
23. The authors thank CNRS and Motorola S.P.S. for
support, V. Collière, L. Datas, and TEMSCAN service
(Université Paul Sabatier Toulouse) for TEM, I. Four-
quaux and B. Payré (Centre de Microscopie Electron-
ique Appliquée à la Biologie. Université Paul Sabatier
Toulouse) for ultramicrotomy, T. Aigouy (Laboratoire
de Mecanismes de Transfert en Geologie, Université
Paul Sabatier Toulouse) for x-ray diffraction, and R.
Andersen (University of California, Berkeley), M. Res-
paud (Laboratoire de Nanophysique, Magnetisme et
Optoelectronique Toulouse), and K. Soulantica (LCC
Toulouse) for fruitful discussions.
Supporting Online Material
www.sciencemag.org/cgi/content/full/303/5659/821/DC1
Figs. S1 to S5
16 October 2003; accepted 22 December 2003

Rheology and Microscopic the time or frequency interval for which the
plateau in the shear relaxation modulus

Topology of Entangled GN(␻) can be observed (1). For natural

rubber (5, 6 ), one can estimate that the
relaxation-time increases from ␶d(M ⫽ 104
Polymeric Liquids g/mol) ⬇ 5 ⫻ 10⫺5 s to ␶d(M ⫽ 106
g/mol) ⬇ 5 s and would reach years for
Ralf Everaers,1,2* Sathish K. Sukumaran,1 Gary S. Grest,3 chains with molecular masses of 108 g/mol
Carsten Svaneborg,1 Arvind Sivasubramanian,1 Kurt Kremer1 even though their typical spatial extension
is only on the order of 1 ␮m.
The viscoelastic properties of high molecular weight polymeric liquids are All melts or semidilute solutions of suf-
dominated by topological constraints on a molecular scale. In a manner ficiently long flexible chain molecules
similar to that of entangled ropes, polymer chains can slide past but not show the same, universal behavior (1). Ear-
through each other. Tube models of polymer dynamics and rheology are ly theories treated entanglements as tran-
based on the idea that entanglements confine a chain to small fluctuations sient physical cross-links to capture the
around a primitive path that follows the coarse-grained chain contour. Here analogy to rubber-elastic polymer net-
we provide a microscopic foundation for these highly successful phenom- works. Modern theories of polymer dynam-
enological models. We analyze the topological state of polymeric liquids in ics are based on the idea that Brownian
terms of primitive paths and obtain parameter-free, quantitative predictions motion is dominated by the restriction that
for the plateau modulus, which agree with experiment for all major classes the chains may slide past but not through
of synthetic polymers. each other. Consequently, the motion of
each polymer chain is thought to be con-
The relation between the complex vis- the plateau shear modulus G 0N, which is on fined to a tubelike region around a so-
coelastic properties of polymer liquids and the order of 106 Pa, or five orders of mag- called primitive path along the coarse-
their microscopic structure and dynamics is nitude smaller than the shear modulus of grained chain contour (7 ) (Fig. 1). The
a key issue in modern materials science and ordinary solids. The origin of this behavior stress relaxation is effectively suspended
biophysics (1–3). A prototypical example is can be traced back to the molecular scale. up to the time ␶d(N) required by the chains
natural rubber. Strictly speaking, natural Natural rubber consists of linear chain mol- to leave or renew their deformed original
rubber is a (highly viscous) liquid, even ecules of up to N ⬇ 3 ⫻ 104 monomer units tubes. The basic version of the tube model
though over a large frequency range it pre- with molecular masses up to M ⫽ 106 considers only a single relaxation mecha-
sents the same rubber-elastic response to g/mol and contour lengths up to L ⫽ 10 nism: one-dimensional, curvilinear diffu-
oscillatory mechanical deformations as ␮m. Individual polymers adopt random coil sion in tubes of fixed length (“reptation”)
vulcanized rubber (4 ). This response can conformations. The coil diameter of about (8). Its prediction ␶d(N) ⬀ N3 agrees with
be characterized by a temperature- and 100 nm is much smaller than the chain experimental data in the limit of large N
concentration-dependent material constant, contour length, but substantially larger than (9). Convincing microscopic evidence for
the diameter of the corresponding densely both tube confinement and reptation dy-
packed globule of ⬃20 nm. Consequently, namics has been accumulated from experi-
1
Max-Planck-Institut für Polymerforschung, Acker- neighboring chains strongly interpenetrate ments and computer simulations (10–12).
mannweg 10, 55128 Mainz, Germany. 2Max-Planck-
Institut für Physik komplexer Systeme, Nöthnitzer Str.
and entangle with each other. In our exam- Recent, more refined analytical (2) and nu-
38, 01187 Dresden, Germany. 3Sandia National Lab- ple, a spherical volume with the coil diam- merical models (13–17 ) of the dynamics of
oratories, Albuquerque, NM 87185, USA. eter contains about 150 polymers. Although the primitive paths account for additional
*To whom correspondence should be addressed. E- the chain length has essentially no effect on relaxation mechanisms and quantitatively
mail: everaers@mpipks-dresden.mpg.de the magnitude of G 0N, it strongly influences describe most rheological and single-chain

www.sciencemag.org SCIENCE VOL 303 6 FEBRUARY 2004 823