ARTICLES

Van der Waals versus dipolar forces controlling mesoscopic organizations of magnetic nanocrystals
Y. LALATONNE, J. RICHARDI AND M. P. PILENI*
Universit Pierre et Marie Curie,Laboratoire des Matriaux Mesoscopiques et Nanomtriques,U.M.R.7070,BP 52,4 place Jussieu,75005 Paris,France *e-mail: pileni@sri.jussieu.fr

Published online: XX xxxx 2003; doi:10.1038/nmat1054

The structure, thermodynamics and dynamics in many physical and chemical systems are determined by interplay of short-range isotropic and long-range anisotropic forces. Magnetic nanoparticles dispersed in solution are ideal model systems to study this interplay, as they are subjected to both isotropic van der Waals and anisotropic dipolar forces. Here we show from experiment an abrupt transition of maghemite nanocrystal organization from chain-like to random structures when nanoparticle solutions are evaporated under a magnetic field. This is explained by brownian dynamics simulations in terms of a variation of the strength of van der Waals interactions with the particle contact distance, which is tuned by the length of the molecules coating the particles. The weak dipoledipole interactions between the maghemite particles are usually not sufficient to result in the chain formation observed here. However, due to the van der Waals interactions, when the nanocrystal contact distance is short enough, clusters of nanocrystals are formed during the evaporation process. These clusters exhibit large dipole moments compared with a single particle, which explains the formation of chain-like structures. Conversely, when the nanocrystal contact distance is too long, no nanocrystal aggregation occurs, and a random distribution of maghemite nanocrystals is obtained.[Author: Some rewording for clarity OK?]
nature materials | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials

anocrystals dispersed in solution are subjected to two major types of attractive forces: van der Waals and dipolar interactions.In the absence of the dipolar interactions, large isotropic van der Waals interactions induce formation of spherical aggregates.For weak van der Waals interactions, the strength of the dipolar moments controls the formation of chain-like structures. This has been well established by simulations for magnetic nanoparticles1-4. Chain-like structures have been directly observed experimentally5. In fact, with no applied magnetic field, it has been shown5 that an abrupt transition from separate particles to randomly oriented chain-like aggregates occurs on increasing the particle size. Similarly, the self-organization of semiconductor CdTe nanocrystals in pearl-necklace aggregates is explained by strong electrical dipole interactions6. The influence of the van der Waals, dipolar and also nanocrystalsubstrate7 interactions on the nanocrystal organization at a mesoscopic scale is a key issue,not only for practical applications but also for fundamental understanding. A variation of the length of the coating molecules[Author: OK?] is used here to tune the interaction between nanocrystals. Maghemite (-Fe2O3) nanocrystals are made by a slight change of the synthesis described previously8,9 (see Methods). The extracted nanocrystals are coated either with octanoic (C7H15COOH) or with dodecanoic acid (C11H23COOH), which are denoted by C8 and C12, respectively. The nanocrystals are then dispersed in cyclohexane.No particle aggregation is observed under small-angle X-ray scattering for diluted dispersions whatever the nanocrystal coating. This confirms the stability of the solutions without any phase separation. The -Fe2O3 nanocrystals, differing by their coatings, are characterized by the same average size (10 nm) and polydispersity (20%). The magnetic properties of these nanocrystals were determined at 3 K by SQUID (superconducting quantum interference device) using a solution with very low volume fraction (0.1%). The same reduced remanence, saturation magnetization and coercive field of 0.30, 77 e.m.u. g1 and 360 Oe, respectively, were measured. This indicates that the nanocrystals, differing only by the chain length of their coating, are characterized by the same magnetic dipolar moments. A transmission electronic microscope (TEM) grid is deposited in the bottom of the beaker containing the solution of nanocrystals dispersed in cyclohexane (1.6 108 mol l1). The system is subjected to a 0.59 T magnetic field, applied parallel to the substrate,during the evaporation.Figure 1 shows drastic changes in the TEM patterns with the various coatings: (i) With no applied magnetic field and a C8 coating, spherical aggregates of various sizes appear (Fig. 1a).[Author: OK? Is this what you meant?] In contrast, the C12-coated nanocrystals are randomly deposited on the
1

ARTICLES
a b c

300 nm

10 m

300 nm

300 nm

10 m

300 nm

Figure 1 TEM images of g-Fe203 nanocrystals deposited with and without a magnetic field. a,Nanocrystals coated with C8 and deposited without a magnetic field form isotropic clusters during the evaporation process. b,The same nanocrystals deposited with a magnetic field organize in chain-like structures. c, Note the presence of isolated nanocrystals between the stripes for the same deposition as in b [Author: OK?]at a higher resolution. d,In contrast,nanocrystals coated with C12 are randomly deposited on the TEM grid in the absence of magnetic field. e,f,Even in the presence of the magnetic field,no organization appears for C12-coated nanocrystals.[Author: OK?]

TEM grid (Fig. 1d). (ii) In a 0.59 T magnetic field, nanocrystals coated with C8 form thick, striped structures in the direction of the applied field (Fig.1b,c).The average length of the chain-like structures is several tens of micrometres, whereas the average diameter is around 300 nm. Note the presence of some nanocrystals between the stripes.Conversely, with C12, the TEM patterns do not show any nanocrystal organization (Fig. 1e,f). A very small number of nanocrystal aggregates seem to be organized along the direction of the applied field. Because the nanocrystals coated with C8 and C12 are characterized by the same magnetic dipolar moments,as shown above,the changes in the organization with the coating shown in Fig. 1 cannot be directly attributed to the magnetic properties of the -Fe2O3 nanocrystals. Usually, high dipolar parameter , defined as the ratio of the magnetic dipolar to thermal energies, are needed for chain-formation of the particles. Assuming that the edgeedge core spacing between the particles is about 1 nm due to their coating, we calculate the dipolar parameter to be 0.69 for the average diameter of 10 nm. Owing to the low size dispersion of the particles, even the largest particles of about 15 nm (2% of all particles) have a value of about 2.5, which is smaller than the threshold , which is 3 for chain formation14.[Author: OK?] Moreover, a simulation study shows that chains of larger dipolar particles do not induce the aggregation of smaller particles in polydisperse systems10. Therefore, the chain formation cannot be directly explained by the presence of larger particles. To explain the formation of the chain-like structures in spite of the weak dipolar interaction, brownian dynamics simulations at a particle level investigate the evaporation of magnetic nanoparticle solutions. The translation and rotation of the magnetic particles are described by Langevin equations11, which make it possible to consider the influence
2

of the solvent on the solute by a combination of random forces and frictional terms. The Langevin equations are integrated using an algorithm proposed by Allen12. The forces and torques between the particles are calculated from the interparticle interaction, which is defined as a sum of the steric repulsion due to the coating molecules,the dipoledipole potential and the van der Waals attraction. We use the interaction model proposed in ref.13,where the explicit equations of all interaction terms are given.The interaction parameters were adapted to the system experimentally used (see interaction and simulation parameters in Methods). The double thickness of the layer of coating molecules, denoted by d in the following, appears as a parameter in the equation for the steric repulsion.By changing d,the influence of the size of the coating molecule on the simulation results can be studied. Simulations were carried out varying d from 1.0 to 3.0 nm. The evaporation process is modelled as follows: a wall at the bottom of the simulation box represents the substrate.The wallparticle interaction is defined by a steric repulsion due to the coating molecules and a van der Waals interaction between a sphere and a half-space14. At the top of the box, there is a liquidgas interface, which slowly moves down during evaporation.[Author: OK?] The interface is made of the solvent and the coating of the particles crossing the interface.When the particle crosses the interface, the interfacial area and, therefore, its energy increases (see explicit formula in Methods).This energy increase confines the particles in the solution layer. The evaporation speed used in the simulations is much faster than the experimental one, which is determined by the boiling point of the solvent. A systematic variation of the evaporation speed in the simulations has shown that the speed used here is sufficiently large to predict qualitatively the morphology of structures. However, the size of the organizations increases at slower evaporation,
nature materials | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials

ARTICLES
a b
1 0 1 Energy (kT) 2 3 4 VdW + steric (d = 1.6 nm) Total potential (d = 1.6 nm) VdW + steric (d = 1.2 nm) Total potential (d = 1.2 nm) 1.1 1.2 1.3 1.4 1.5

5 6

Particleparticle distance (/ 10 nm)

Figure 3 Dependence of the interaction energy on the interparticle distance. The total energy and the sum of the van der Waals attraction and the steric repulsion are shown for two values of the coating layer thickness d.For the calculation of the dipolar interaction,the most stable headtail configuration is chosen.The minimum of the total energy decreases from 3.0 to 4.5 kT on reducing d from 1.6 to 1.2 nm due to the sensitivity of the van der Waals potential to the interparticle distance.[Author: Please can you clarify the scale on the x axis,do you mean that the distances should be multiplied or divided by 10,i.e.that they are 0.11 or 11.0 etc? Best to label them at the actual value?]

Figure 2 Snapshots of the configurations by brownian dynamics simulations. a,b,For the coating layer thickness d = 1.2 nm before evaporation without (a) or with (b) an applied field,no significant aggregation of particles is observed. c,d,After evaporation, without applied field (c),spherical clusters of particles are observed; whereas with an applied field (d) elongated clusters of particles appear. e,f, For a thicker coating (d = 1.6 nm),no particle aggregation is observed even during the evaporation when particles are subjected (f) or not (e) to an applied magnetic field. [Author: Some rewording of caption for clarity,OK?]

which may explain the differences in the size of aggregates observed by simulations and experiments. Furthermore, the use of a very low evaporation speed in simulations leads to qualitatively the same mesostructures, but the nanoparticles are well organized in a 3D network. The polydispersity of the -Fe2O3 nanocrystals is not low enough to observe this internal organization experimentally. All simulations are started from a configuration obtained for d = 3.0 nm,where no aggregation of particles occurs.Figure 2a,b shows the configuration after an equilibration of ten million time steps for a coating layer of 1.2 nm. No significant aggregation of particles is observed when they are (Fig. 2b) or not (Fig. 2a) subjected to a magnetic field (0.59 T). For d = 1.2 nm and 1.6 nm, similar configurations were obtained. The major changes are observed during the evaporation process: (i) For d = 1.2 nm, the particles are aggregated after 14 million steps. With no applied magnetic field, spherical clusters of particles are
nature materials | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials

observed (Fig. 2c) whereas, in a field, alignments of particles are found (Fig. 2d).The aggregation observed for d = 1.2 nm (Fig. 2c) is explained by a larger interparticle attraction.In fact,Fig. 3 shows that the energy of the most stable head-tail configuration decreases from 3.0 to 4.5 kT on reducing d from 1.6 to 1.2 nm. This large variation is mainly due to the marked sensitivity of the van der Waals potential to the interparticle distance.The stronger attraction leads to a condensation of the particles. To explain the formation of striped structures with -Fe2O3 nanocrystals in a magnetic field in spite of a low dipole interaction with respect to the thermal energy (Fig. 2d), both van der Waals attraction and magnetic dipoles have to be taken into account:when the coating molecule is small enough clustersare formed due to the van der Waals forces. The cluster formation considerably enhances the long-range dipolar forces compared with isolated particles with creation of a macro dipolar momentleading to an anisotropic organization of the nanoparticles. It is interesting to note that similar behaviour was observed for a Stockmayer fluid on using Monte-Carlo simulations15. In a field, an elongation and coalescence of droplets made of dipolar particles appear. However this was observed for a larger dipole parameter ( = 2.8) as in our case. (ii) For d = 1.6 nm, no aggregation of the particles is observed even during the evaporation when particles are subjected (Fig. 2f) or not (Fig. 2e) to an applied magnetic field (0.59 T) during the evaporation process.The minima of the potentials in Fig. 3 indicate that the distance of closest contact increases from 1.2 to 1.6 nm. For the contact distance of 1.6 nm, the van der Waals interactions are not strong enough to induce formation of particle clusters, preventing any enhancement of the dipolar forces for such an assembly of particles. Rather good agreement between the structures obtained from experiments (Fig. 1) and those obtained from simulations (Fig. 2) can be observed:(i) With -Fe2O3 nanocrystals coated with C8,large clusters are observed from the TEM pattern (Fig. 1a). This structure is in qualitative agreement with the particle aggregation found by
3

ARTICLES
a d

2 m

2 m

10 m

10 m

2 m

2 m

Figure 4 SEM images of -Fe203 nanocrystals. ac,Nanocrystals coated with C8: a, deposited without a magnetic field,form spherical aggregates; b,with an applied field, elongated structures are formed; and c,by tilting the sample,superimposed tubes are observed. df,For nanocrystals coated with C12 and deposited without applied field,a very dense film is observed (d);[Author: The scale bar is missing from part d,should it be the same as on part a?] e, when a magnetic field is applied during the evaporation process,some elongated structures appear,but by tilting the sample (f) only slight undulations on the surface of the thick film are observed. [Author: Some rewording of caption for clarity,OK?]

observed in this system, when octanthiol (C8H17SH) is replaced by dodecanthiol (C12H25SH) as the coating agent. On the one hand, these experimentally observed spacings are in reasonable agreement with the values used here in the simulations. On the other hand, the increase of 0.3 nm in spacing experimentally expected between C8 and C12 is similar to the small variation of 0.4 nm necessary for the disappearance of organization in simulations. The change in the nanostructure with the coating shown above should be also found at higher particle concentration. The same experimental procedure as described above has been used. The TEM grid is replaced by HOPG (high-oriented pyrolitic graphite) and the nanocrystal concentration is 6.1 106 mol l1 instead of 1.6 108 mol l1. The mesoscopic structures are now observed by scanning electron microscopy (SEM). Figure 4a shows a very rough thick film of highly compact spherical aggregates with an average diameter of 1.5 m for C8 nanocrystals. In a magnetic field (0.59 T), the mesoscopic structure markedly changes with formation of long bands with a very regular structure (Fig. 4b). By tilting the sample, superimposed tubes with an average diameter of 3 m are observed (Fig. 4c).[Author:OK?] These data are consistent with those obtained in dilute solution and explained by cluster formation. At high nanocrystal concentration, the number of clusters increases with formation of large spherical aggregates as observed in Fig. 4a. In an applied magnetic field, the clusters formed induce large dipolar forces allowing the particles to form cylinders. Because of the high particle concentration, the cylinders tend to superimpose with a diameter ten times larger than that observed in dilute solution. With C12 nanocrystals, as expected from above, the film behaviour markedly differs from that of the C8 derivative. Because in dilute solution the nanocrystals are randomly dispersed (Fig. 1d), at high nanocrystal concentration a thick, compact film with a very flat surface and is observed (Fig. 4d).[Author: OK?] In a 0.59-T magnetic field applied during the evaporation process, slight undulations on the surface of the thick film are observed (Fig. 4f). On a macroscopic scale the film does not change (Fig. 4e) compared with that obtained with no applied magnetic field (Fig. 4d). This is in good agreement with results obtained in dilute solution either experimentally or by simulation. From the data described above, it is concluded that elongated assemblies are obtained even for weakly dipolar particles by a careful choice of the coating agent. As the origin of these structures is quite different from that in highly dipolar fluids,the structural and dynamical study can give fundamental insights into dipolar fluids in general.The Fe2O3 nanocrystals provide model systems to explore the influence of other interaction terms, such as hydrophobic attraction or the electrostatic repulsion between charged nanoparticles, on the nanoparticle organizations.
METHODS

simulations for d = 1.2 nm (Fig. 2c). Similarly, the TEM images show a striped structure when the nanocrystals are subjected to a magnetic field during the deposition process (Fig. 1b,c) whereas aligned particles are seen by simulations (Fig. 2d). (ii) With -Fe2O3 nanocrystals coated with C12,random distributions of particles are observed from the TEM pattern (Fig. 1d) as obtained by simulation for d = 1.6 nm (Fig. 2e). When the system is subjected to a magnetic field during the deposition process, no changes in the nanocrystals dispersion are observed, either experimentally (Fig. 1e,f) or by simulation (Fig. 2f). In simulations,the aggregation disappears when the edgeedge core spacing between the particles is varied from 1.2 to 1.6 nm. A similar disappearance of organization is experimentally observed when C12 instead of C8 is used as the coating. An idea of the interparticle spacing that can be expected for these coatings, is obtained from Ag2S nanocrystals coated with alkanethiols and self-assembled in hexagonal networks16. An increase in the core spacing from 1.4 to 1.7 nm is
4

SYNTHESIS OF SOLUTIONS OF MAGHEMITE NANOCRYSTALS

To prepare non-coated -Fe2O3 particles, the first step is to add dimethylamine ((CH3)2NH2OH) to an aqueous micellar solution of ferrous dodecyl sulphate (Fe(DS)2)17. The final concentrations after mixing of the reactants are 1.3 102 mol l1 and 8.5 10-1 mol l1 for Fe(DS)2 and dimethylamine, respectively. The solution is stirred vigorously for 2 h at 28.5 C and the resulting precipitate of uncoated nanocrystals is isolated from the supernatant by centrifugation. In the second step, this precipitate is washed with a large excess of ethanol and a solution of carboxylic acid solubilized in ethanol ([CnH2n+1COOH]= 1.4 101 mol l1) is added. The solution is subjected to sonication for two hours at 90 C and the precipitate that appears is washed with a large excess of ethanol and the powder is dried in air. The nanocrystals coated with octanoic acid (C8) or dodecanoic acid (C12) are dispersed in cyclohexane. All the solutions are kept on a magnet for 12 hours in order to discard unstable nanoparticles and the supernatant is then collected. X-ray diffraction lines of deposited nanocrystals indicate an inverted spinel phase with a lattice constant of 8.36 , which is characteristic for -Fe2O3. The TEM images of deposited nanocrystals indicate that the average diameter (10 nm) and the polydispersity (20%) of the nanocrystals remain the same, whatever the coating is. (All the chemicals were bought from Aldrich or Fluka).

PREPARATION AND CHARACTERIZATION OF THE -FE2O3 NANOCRYSTALS DEPOSITIONS

The nanocrystals are deposited on a TEM grid by evaporation of a ferrofluid solution (8 hours). During this process, a magnetic field is applied parallel to the substrate plane. A TEM grid with an amorphous

nature materials | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials

ARTICLES
carbon layer is dipped into 200 l of a solution with a nanocrystal concentration of 1.6 108 mol l1. We have verified that the same amount of nanoparticles (0.45 g) is always deposited. For higher particle concentration, the same experimental procedure as described above has been used. An HOPG substrate is dipped into 200 l of a solution with a nanocrystal concentration of 6.2 106 mol l1. A Jeol (100 kV) model JEM 100 CX II TEM was used to obtain micrographs of the ferrite particles. X-ray diffraction measurements[Author: Please give make and model of equipment used.] were carried out with a rotating anode generator (copper anode; focus size: 0.2 0.2 mm2, 50 kV, 30 mA). A Jeol model JSM-840A SEM was used to visualize the assemblies of ferrite nanocrystals. Hysteresis loops at 3 K were measured with a commercial SQUID magnetometer with fields up to 5 T.[Author: Please give make and model of SQUID.] 4. Morimoto, H. & Maekawa, T. Dynamic analysis of a ferromagnetic colloidal system. Int. J. Mod. Phys. B 13, 20852092 (1999). 5. Butter, K., Bomans, P. H. H., Frederik, P. M., Vroege, G. J. & Philipse, A. P. Direct observation of dipolar chains in iron ferrofluids by cryogenic electron microscopy. Nature Mater. 2, 8891 (2003). 6. Tang, Z., Kotov, N. A. & Giersig, M. Spontaneous organization of single CdTe nanoparticles into luminescent nanowires. Science 297, 237240 (2002). 7. Motte, L., Lacaze, E., Maillard, M. & Pileni, M. P. Self-assemblies of silver sulfide nanocrystals on various substrates. Langmuir 16, 38033812 (2000). 8. Ngo, A. T. & Pileni, M. P. Nanoparticles of cobalt ferrite: Influence of the applied field on the organization of the nanocrystals on a substrate and on their magnetic properties. Adv. Mater. 12, 276279 (2000). 9. Ngo, A. T. & Pileni, M. P. Assemblies of ferrite nanocrystals: Partial orientation of the easy magnetic axes. J. Phys. Chem B 105, 5358 (2001). 10. Wang, Z. & Holm, C. Structure and magnetic properties of polydisperse ferrofluids: a molecular dynamics study. Phys. Rev. E 68, 041401 (2003). 11. Allen, M. P. & Tildesley, D. J. Computer Simulations of Liquids (Oxford Science, Oxford 1987). 12. Allen, M. P. Brownian dynamics simulation of chemical reaction in solution. Mol. Phys. 40, 10731079 (1980). 13. Morimoto, H. & Maekawa, T. Cluster Structures and cluster-cluster aggregations in a two dimensional colloidal system. J. Phys. A 33, 247258 (2000). 14. Hunter, R. J. Foundations of Colloid Science (Oxford Univ. Press, Oxford 2001) 15. Stevens, M. J. & Grest, G. S., Phase coexistence of a Stockmayer fluid in an applied field. Phys. Rev. E 51, 59765984 (1995). 16. Motte, L. & Pileni, M. P. Influence on the 2D and 3D self organization of the length of the alkyl chains used to passivate silver sulfide nanoparticles. J. Phys. Chem. B 102, 41044109 (1998). 17. Moumen, N., Veillet, P. & Pileni, M. P. Controlled preparation of nanosize cobalt ferrite magnetic particles. J. Magn. Magn. Mater. 149, 6771 (1995). 18. Rosensweig, R. E. Ferrohydrodynamics (Cambridge Univ. Press, Cambridge 1985).

PARAMETERS OF THE INTERACTION MODEL AND OF THE SIMULATIONS

The magnetic moment is calculated from the -Fe203 bulk magnetization (3.73 105 Am1) and the average size of the particle (10 nm). A surface density of 5 1018 mol m2 for the surfactant molecules is used to calculate the steric repulsion. The Hamaker constant A of the ferrite particles, 1019 J, is taken from the literature4,18. The friction coefficients are calculated from Stokes law using a viscosity of 103 Kg m1 s1 (ref. 4). A small Hamaker constant of 1020 J is chosen for the particlewall interaction to avoid a too large attraction between particle and substrate. For the gas-coating and gas-solvent interfacial tensions, we used the same value of = 18 103 J m2, which is characteristic of organic solvents17. We assume that the energy increase for a particle i crossing the gas-solution interface can be calculated from the equation: 2 2 ui = 2ap(ap dgp) + (ap dgp ), where ap = (dp + d)/2 > dgp; dgp is the distance between the centre of the particle and the gassolution interface and dp is the particle diameter. The first term corresponds to the energy due to the interface between the particle coating and the gas phase, whereas the second term is caused by the decrease in the solventgas interface due to the penetration of the interface by the particle. The numbers of particles used for the simulations are 125 and 512 (shown here) and the initial box lengths are fixed at 160 and 320 nm, respectively. Insignificant differences between 125 and 512 particle simulations are found. The time step for the simulations is 20 ps.

Received 19 August 2003; accepted 1 December 2003; published XX xxxxx 2003. References
1. Chantrell, R. W. Bradbury, A., Popplewell, J. & Charles, S. W. Agglomerate formation in a magnetic fluid. J. Appl. Phys. 53, 27422744 (1982). 2. Tlusty, T. & Safran, S. A. Defect-induced phase separation in dipolar fluids. Science 290, 13281331 (2000). 3. Tavares, J. M., Weiss, J. J. & Telo da Gama, M. M. Strongly dipolar fluids at low densities compared to living polymers. Phys. Rev. E 59, 43884395 (1999).

Acknowledgements
The authors thank E. Vincent for giving them the opportunity to use the SQUID equipment of CEA Saclay[Author: Please spell out CEA.], A. T. Ngo for providing the know-how to make the nanocrystals, and C. Petit for helping Y. Lalatonne with the SAXS experiments. Special thanks are due to L. Motte for supervising Y. Lalatonne in the synthesis and preparation of the 3D superstructures. Correspondence and requests for materials should be addressed to M.P.P.

Competing financial interests

The authors declare that they have no competing financial interests.

nature materials | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials