LETTERS

Binary nanocrystal superlattice membranesself-assembled at the liquid–air interfaceAngang Dong1, Jun Chen2, Patrick M. Vora3, James M. Kikkawa3 & Christopher B. Murray1,2

The spontaneous organization of multicomponent micrometre-sized colloids1 or nanocrystals2 into superlattices is of scientificimportance for understanding the assembly process on the nano-metre scale and is of great interest for bottom-up fabrication offunctional devices. In particular, co-assembly of two types ofnanocrystal into binary nanocrystal superlattices (BNSLs) hasrecently attracted significant attention2–8, as this provides a low-cost, programmable way to design metamaterials4 with preciselycontrolled properties that arise from the organization and inter-actions of the constituent nanocrystal components9. Althoughchallenging, the ability to grow and manipulate large-scale BNSLsis critical for extensive exploration of this new class of material.Here we report a general method of growing centimetre-scale,uniform membranes of BNSLs that can readily be transferred toarbitrary substrates. Our method is based on the liquid–air inter-facial assembly of multicomponent nanocrystals and circumventsthe limitations associated with the current assembly strategies,allowing integration of BNSLs on any substrate for the fabrica-tion of nanocrystal-based devices10. We demonstrate the construc-tion of magnetoresistive devices by incorporating large-area(1.5 mm 3 2.5 mm) BNSL membranes; their magnetotransportmeasurements clearly show that device magnetoresistance isdependent on the structure (stoichiometry) of the BNSLs. The abilityto transfer BNSLs also allows the construction of free-standingmembranes and other complex architectures that have not beenaccessible previously.

In comparison with the growth of single-component superlattices11,the assembly process for BNSLs is more complicated, with a number ofdistinct pairwise interactions (van der Waals, dipolar, Coulombic andso on) combining with hard-sphere space-filling rules to drive the self-assembly of multicomponent nanocrystals9. Despite the recent advancesin the growth and structural characterization of BNSLs12–17, their appli-cations are limited by the fact that the formation and manipulation oflarge-scale BNSLs remain a major challenge14. Existing methods ofgrowing BNSLs rely on the evaporation of a binary nanocrystal solutionon a tilted solid substrate2,4,12–17. In this approach, BNSL formation isinduced by the evaporation of high-boiling-point solvents such as tetra-chloroethylene, such that the assembly process typically takes severalhours to complete2,4,12–17. In addition, the substrate wetting propertieshave significant influence on BNSL formation. Until now, growth ofBNSLs has been limited to the use of hydrophobic substrates such ascarbon-coated transmission electron microscopy (TEM) grids12,making it difficult to integrate BNSLs on a desired substrate for deviceapplications. Extended BNSL growth also requires careful regulation ofthe solvent evaporation kinetics. To obtain BNSLs with relatively largedomain sizes, the assembly process is usually performed under reducedpressure at elevated temperatures (that is, 45–75 uC)2,4,12–17. However,even under optimized conditions, the as-grown BNSLs typically exist as

isolated, micrometre-sized islands scattered irregularly on the substrate,leading to a low surface coverage14,15. Our method is based on drying-mediated self-assembly of multicomponent nanocrystals on an immis-cible liquid surface, allowing the growth of macroscopic BNSLmembranes with no substrate restrictions under ambient conditions.In this procedure, hexane replaces high-boiling-point solvents, allowingthe assembly process to complete in several minutes. Another majorbenefit of our approach is that BNSL membranes formed on the liquidsurface can be readily transferred to arbitrary substrates.

The Fe3O4–FePt nanocrystal combination is a model system withwhich to demonstrate the formation of BNSL membranes, as thecontrast difference between two nanocrystal components in TEMallows better observation of the superlattice structure. To growBNSL membranes, we spread a drop of hexane solution containingFe3O4 and FePt nanocrystals with selected sizes and concentrationsover the surface (,1.5 cm 3 1.5 cm) of diethylene glycol (DEG) in aTeflon well (Fig. 1a). The well was covered with a glass slide andhexane was then allowed to evaporate over 5–10 min, resulting in asolid film supported on the DEG subphase surface. To transferBNSLs, a substrate was placed under the floating film and then gentlylifted up. In this way, centimetre-scale BNSL membranes can bereadily transferred to the substrate after drying. The photograph inFig. 1a shows a SiO2–Si wafer (,7 mm 3 7 mm) that was coated by aBNSL membrane self-assembled from 15-nm Fe3O4 and 6-nm FePtnanocrystals (,3:1 FePt/Fe3O4 particle ratio). TEM showed that themembrane is continuous and uniform in thickness, consisting of amosaic of AlB2-type BNSL domains (Fig. 1b–g) with different growthdirections or different lattice plane projections, as evidenced by thegrain boundaries between neighbouring domains (SupplementaryFig. 1). The average domain size is 1–2mm, with some large domainsapproaching tens of micrometres in lateral dimension. Both the (001)and the (100) lattice projection were commonly observed in TEM(Fig. 1b, e) as well as in high-resolution scanning electron microscopy(Fig. 1d, g). In some cases, superlattice domains isostructural withCu3Au and icosahedral-AB13 (ico-AB13) structures were also observedas minority phases coexisting with the AlB2-type superlattices (Sup-plementary Fig. 2). Simultaneous formation of different types ofBNSLs has also been observed previously2,12, and is presumably attri-butable to the local variations in nanocrystal concentration duringBNSL growth.

We used atomic force microscopy (AFM) to characterize the thick-ness and surface morphology of BNSL membranes transferred toSiO2–Si wafers. The membrane thickness was found to range between30 and 60 nm (depending on the sample), which corresponds to 1–3unit cells. Figure 2a shows the AFM height image of a typical AlB2-type BNSL membrane, the thickness of which is ,40 nm as deter-mined from the line profile analysis (Fig. 2a, inset). The AFM phaseimage revealed a (100) lattice projection in plane view (Fig. 2b),

1Department of Chemistry, University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA. 2Department of Materials Science and Engineering, University of Pennsylvania,Philadelphia, Pennsylvania 19104, USA. 3Department of Physics & Astronomy, University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA.

Vol 466 | 22 July 2010 | doi:10.1038/nature09188

474Macmillan Publishers Limited. All rights reserved©2010

suggesting a (001) projection from a certain side view. Given thepacking geometry of the (001) plane and the sizes of the two nano-crystal components, the membrane thickness is about 1 unit cell. Themembrane is comprised of two layers of large Fe3O4 nanocrystals andfour layers of small FePt nanocrystals (Fig. 2c). In addition, the root-mean-squared roughness value determined from a 10mm 3 10 mmscan window containing multiple domains is about 5 nm, indicatingthat the membrane surface is flat and smooth. Individual domainswere found to vary by 61 nanocrystal layer in thickness, furtherconfirming that the membrane is of uniform thickness. These resultsare consistent with scanning electron microscopy (SEM) studies(Supplementary Fig. 3).

The structure and stoichiometry of BNSL membranes can be tunedby changing the size and/or the concentration ratios between the two

nanocrystal components. For example, NaCl-type BNSL membranes(Supplementary Fig. 4) were formed by co-crystallization of 15-nmFe3O4 and 4-nm FePt nanocrystals (,2:1 FePt/Fe3O4 particle ratio).Notably, sharp cracks propagating in the membrane were observedupon fracture (Supplementary Fig. 4c, d), further confirming thelong-range nanocrystal ordering throughout the membrane.Further increasing the FePt/Fe3O4 particle ratio resulted in CaB6-typeBNSL membranes (Supplementary Fig. 5). In addition to the Fe3O4–FePt combination, BNSL membranes with tunable structures wereobtained in the assembly of a variety of binary nanocrystal systemssuch as Fe3O4–Fe3O4 and Fe3O4–CoPt3 (Supplementary Fig. 6), indi-cating the generality of the approach.

Growing BNSLs on a liquid surface facilitates film transfer, allow-ing integration of BNSLs on any substrate for device fabrication. Webuilt magnetoresistive devices (Fig. 3) by incorporating large-area(1.5 mm 3 2.5 mm) BNSL membranes self-assembled from 15-nmFe3O4 and 7-nm Fe3O4 nanocrystals. The spin-dependent transportof Fe3O4 nanocrystal arrays has recently attracted considerable atten-tion18,19, as Fe3O4 could potentially achieve high magnetoresistance(MR 5 (R 2 R0)/R0, where R is the resistance and R0 is its zero-fieldvalue) owing to its high spin polarization20. Co-assembly of twodifferent-sized Fe3O4 nanocrystals into BNSLs provides an oppor-tunity to precisely tune magnetoresistance by varying the superlatticestructure (stoichiometry). To fabricate the device, a floating AlB2-type (Supplementary Fig. 6a) or ico-AB13-type (Supplementary Fig.6b) BNSL membrane was transferred to a sapphire substrate prepat-terned with gold electrodes (Fig. 3a) and then annealed at 500 uC toreduce the interparticle spacing18. TEM showed that the binarysuperlattice structure was preserved after annealing (Fig. 3b, c). Forboth types of BNSL, the membrane resistivity increased mono-tonically with decreasing temperature (T) and the zero-bias-voltageconductance (G0) as a function of T can be well fitted by ln(G0) /T21/2 (Fig. 3d). This suggests that some form of thermally assistedinterparticle tunnelling in the presence of electron–electron interac-tions, such as variable range hopping or co-tunnelling, dominates theelectronic transport through BNSL membranes18,19,21.

Measurements of device resistance as a function of the externalmagnetic field (H) yielded negative magnetoresistance, with the mag-netoresistance values gradually increasing with decreasing T (Fig. 3e).For each type of superlattice structure, we measured two distinctmembranes and four devices were fabricated from each membrane.The magnetoresistance values obtained are consistent from device to

DEG

Cover

cm

a

100 nm

100 nm

b c

d

f

g

e

Figure 1 | Large-scale BNSL membranes self-assembled at the liquid–airinterface. a, Schematic of the BNSL membrane growth and transferprocesses. The photograph shows a typical BNSL membrane transferred to aSiO2–Si wafer. Mechanical damage from tweezers in the membrane’s upperright corner (photo) helps visualize the scale. b–g, AlB2-type BNSLmembranes self-assembled from 15-nm Fe3O4 and 6-nm FePt nanocrystals.b, TEM image of the (001) lattice projection (upper inset, magnified view;lower inset, small-angle electron diffraction pattern). c, d, Crystallographicmodel (c) and high-resolution scanning electron microscopy (HRSEM;d) image of the (001) projection. e, TEM image of the (100) lattice projection(upper inset, magnified view; lower inset, small-angle electron diffractionpattern). f, g, Crystallographic model (f) and HRSEM image of the (100)projection (g).

0.8

0.6

0.4

0.2

0 0°

10°

20°

30°

40°

b

(001)(001)

(100)(100)

0

10

20

30

40

50

60

70

0

0.8

0.6

0.4

0.2

0(μm)

BNSL membrane SiO2–Si wafer

a

40 nm

c

40

20

00 200

0.80.60.40.2 0 (μm)0.80.60.40.2

(μm) (μm)

800600400 (nm)

(nm)

Figure 2 | BNSL membrane one unit cell thick. a, AFM height image (scansize, 1 mm 3 1mm) of an AlB2-type BNSL membrane consisting of 15-nmFe3O4 and 6-nm FePt nanocrystals. Inset, height analysis of the profileindicated in the AFM image. b, AFM phase image of the same membrane,showing a (100) projection in plane view. c, Side view of a crystallographicmodel of the membrane, showing that the membrane is one unit cell thick.

NATURE | Vol 466 | 22 July 2010 LETTERS

475Macmillan Publishers Limited. All rights reserved©2010

device (magnetoresistance difference, ,0.02%), suggesting the highreproducibility of devices, which we attribute to the high uniformityof the BNSL membranes. This allowed us to compare magnetotran-sport between two types of binary superlattice, which clearly showedthat the AlB2-type BNSL membranes had higher magnetoresistancevalues than the ico-AB13-type membranes, with the magnetoresis-tance difference increasing with decreasing T (Fig. 3f). Because thesetwo types of BNSL were assembled from the same nanocrystal com-ponents, the difference in their superlattice structures (stoichi-ometry) is believed to account for the observed magnetoresistancedifference. Further work is underway to study the structure-dependentmagnetoresistance of BNSLs systematically. It is also of interest toexplore the magnetotransport of BNSL membranes assembled fromFe3O4 and other magnetic nanocrystals such as FePt and CoPt3.

The ability to transfer BNSLs also allows the construction of uniquearchitectures such as free-standing membranes, which are ideal struc-tures with which to fabricate devices free of substrate-induced inter-ference22,23. Existing methods do not allow the growth of free-standingmembranes of BNSLs. In contrast, by simply transferring a floatingfilm to a holey TEM grid and then drying it, we are able to fabricatefree-standing BNSL membranes with areas of 7mm 3 7mm (Fig. 4).Occasionally, fractured regions help visualize the membranes sus-pended in the holes (Fig. 4a, c). Most free-standing membranes(,90%) remained intact through transfer and vacuum drying(Fig. 4a), indicating that these membranes possess considerable mech-anical robustness, which is presumably attributable to their long-range-ordered superlattice structure. In addition, sequential transferof the floating membranes produced multilayer BNSL films, furtherdemonstrating the flexibility of this approach. Such multilayer filmscan exhibit interesting Moire patterns24 (Supplementary Fig. 7), arisingfrom the misorientation between two layers with respect to each other.

We found that the solvent evaporation kinetics and the subphaseproperties had substantial influence on the growth of BNSL mem-branes. Replacing hexane with higher-boiling-point solvents such astoluene led to random mixtures of two nanocrystal components.Self-assembly of binary nanocrystals on other immiscible liquids

1.5 mm

2.5

mm

20 nm 20 nm

AlB2 ico-AB13

a

b c

f

d

e

300 K

200 K

100 K

–10

–12

–14

–16

–18

–20

–220.05 0.06 0.07 0.08 0.09 0.10 0.11

0

–1

–2

–3

–4

–5–1.0 –0.5 0.0 0.5 1.0

H (T)

–1.5

–2.0

–2.5

–3.0

–3.5

–4.0

–4.5

100 150 200 250 300

T (K)

T–1/2 (K–1/2)

MR

(%)

MR

(%)

Ln(G

0)

Figure 3 | Magnetotransport of large-area (1.5 mm 3 2.5 mm) BNSLmembranes. a, Schematic of the device. b, TEM image of the thermallyannealed AlB2-type BNSL membrane. c, TEM image of the thermallyannealed ico-AB13-type BNSL membrane. d, Temperature dependence ofzero-bias-voltage conductance (G0) for the AlB2-type (red) and ico-AB13-type (blue) BNSL membranes. G0 is plotted on a logarithmic scale. The black

lines are the corresponding linear fits to the experimental data.e, Magnetoresistance of AlB2-type (red) and ico-AB13-type (blue) BNSLmembranes at various temperatures. f, Temperature dependence ofmagnetoresistance for the AlB2-type (red) and ico-AB13-type (blue) BNSLmembranes at H 5 1 T.

20 nm100 nm

a b

c d

5 μm 50 nm

Figure 4 | Free-standing BNSL membranes consisting of 11-nm Fe3O4 and4.5-nm FePt nanocrystals. a, TEM overview shows that most free-standingmembranes suspended in the holes remained intact through transfer anddrying. One fractured membrane in the upper right corner helps visualizethe membrane. b, Magnified view of the region indicated in a, showing themembrane edge attached to the copper grid as well as the AlB2-type structureof the membrane. c, TEM image of a fractured free-standing membrane,showing the sharp crack along the fracture. d, Magnified view of the regionindicated in c.

LETTERS NATURE | Vol 466 | 22 July 2010

476Macmillan Publishers Limited. All rights reserved©2010

such as water resulted in phase separation or films with limited BNSLstructures, whereas continuous BNSL membranes could be repro-ducibly grown on DEG or tetraethylene glycol surfaces under thesame experimental conditions. DEG has previously been used asthe subphase to prepare monolayer nanocrystal arrays by theLangmuir–Blodgett method25,26. We attributed DEG utility in thegrowth of BNSLs to its chemical inertness9 and low evaporation rate.When ethylene glycol was used as the subphase, BNSL islands, insteadof continuous films, were observed (Supplementary Fig. 8).

We have demonstrated that co-crystallization of multicomponentnanocrystals at the liquid–air interface is a simple yet robustapproach to rapidly growing large-scale membranes of long-range-ordered BNSLs. The ability to transfer BNSLs allows the fabricationof nanocrystal-based devices, which will accelerate the exploration ofthis new and important class of material. Given that this assemblystrategy is applicable to different nanocrystal combinations, weanticipate that membranes of quasicrystalline BNSLs17 and ternarynanocrystal superlattices27 will also be grown by this method.

METHODS SUMMARYNanocrystal synthesis. We synthesized monodisperse Fe3O4 (ref. 28), FePt (ref. 29)

and CoPt3 (ref. 30) nanocrystals with tunable sizes according to the literature

methods. As-synthesized nanocrystals were purified by precipitation and were dis-

persed in hexane to form stable solutions with concentrations of 1–10 mg ml21.

Self-assembly and transfer of BNSL membranes. In a typical assembly process

for an AlB2-type membrane, we mixed a 10-ml hexane solution of 15-nm Fe3O4

nanocrystals (,6 mg ml21) with a 10-ml hexane solution of 6-nm FePt nano-

crystals (,3 mg ml21) in a vial. The mixed solution was spread on the DEG

surface in a Teflon well. The well was covered with a glass slide and hexane

was then allowed to evaporate through the gaps between the Teflon surface

and the glass slide (evaporation rate, 3–5ml min21). A solid membrane formed

upon complete evaporation of hexane. To transfer BNSLs, a substrate was placed

under the floating membrane using tweezers and then gently lifted upwards to

collect the membrane. The subsequent drying in a vacuum chamber removedresidual DEG.

Instrumentation. For TEM and SEM characterization, BNSL membranes were

collected on carbon-coated copper grids or holey copper grids (hole size,

7mm 3 7mm). TEM images and small-angle electron diffraction patterns were

obtained using a JEOL-1400 TEM operating at 120 kV, and SEM images were

obtained using a JEOL 7500F SEM operating at 2 kV. AFM was carried out on a

Digital Instruments multimode scanning probe microscope. For magnetotransport

studies, BNSL membranes were transferred to sapphire substrates prepatterned

with Cr–Au electrodes (channel length, 1.5 mm; channel width, 2.5 mm) by ther-

mal evaporation. The device was then annealed at 500 uC under vacuum (1023 torr)

for 1 h. Magnetotransport measurements were performed on a Physical Property

Measurement System from Quantum Design. Structural models were built using

Materials Studio 4.4 (Accelrys Software).

Received 25 January; accepted 20 May 2010.

1. Leunissen, M. E. et al. Ionic colloidal crystals of oppositely charged particles.Nature 437, 235–240 (2005).

2. Shevchenko, E. V., Talapin, D. V., Kotov, N. A., O’Brien, S. & Murray, C. B.Structural diversity in binary nanoparticle superlattices. Nature 439, 55–59(2006).

3. Kiely, C. J., Fink, J., Brust, M., Bethell, D. & Schiffrin, D. J. Spontaneous ordering ofbimodal ensembles of nanoscopic gold clusters. Nature 396, 444–446 (1998).

4. Redl, F. X., Cho, K. S., Murray, C. B. & O’Brien, S. Three-dimensional binarysuperlattices of magnetic nanocrystals and semiconductor quantum dots. Nature423, 968–971 (2003).

5. Saunders, A. E. & Korgel, B. A. Observation of an AB phase in bidispersenanocrystal superlattices. ChemPhysChem 6, 61–65 (2005).

6. Cheon, J. et al. Magnetic superlattices and their nanoscale phase transitioneffects. Proc. Natl Acad. Sci. USA 103, 3023–3027 (2006).

7. Kalsin, A. M. et al. Electrostatic self-assembly of binary nanoparticle crystals witha diamond-like lattice. Science 312, 420–424 (2006).

8. Zheng, J. et al. Two-dimensional nanoparticle arrays show the organizationalpower of robust DNA motifs. Nano Lett. 6, 1502–1504 (2006).

9. Talapin, D. V. LEGO materials. ACS Nano 2, 1097–1100 (2008).10. Urban, J. J., Talapin, D. V., Shevchenko, E. V., Kagan, C. R. & Murray, C. B.

Synergism in binary nanocrystal superlattices leads to enhanced p-type

conductivity in self-assembled PbTe/Ag2Te thin films. Nature Mater. 6, 115–121(2007).

11. Murray, C. B., Kagan, C. R. & Bawendi, M. G. Synthesis and characterization ofmonodisperse nanocrystals and close-packed nanocrystal assemblies. Annu. Rev.Mater. Sci. 30, 545–610 (2000).

12. Shevchenko, E. V., Talapin, D. V., Murray, C. B. & O’Brien, S. Structuralcharacterization of self-assembled multifunctional binary nanoparticlesuperlattices. J. Am. Chem. Soc. 128, 3620–3637 (2006).

13. Chen, Z. Y., Moore, J., Radtke, G., Sirringhaus, H. & O’Brien, S. Binary nanoparticlesuperlattices in the semiconductor-semiconductor system: CdTe and CdSe. J.Am. Chem. Soc. 129, 15702–15709 (2007).

14. Smith, D. K., Goodfellow, B., Smilgies, D. M. & Korgel, B. A. Self-assembled simplehexagonal AB2 binary nanocrystal superlattices: SEM, GISAXS, and defects. J. Am.Chem. Soc. 131, 3281–3290 (2009).

15. Overgaag, K. et al. Binary superlattices of PbSe and CdSe nanocrystals. J. Am.Chem. Soc. 130, 7833–7835 (2008).

16. Friedrich, H. et al. Quantitative structural analysis of binary nanocrystalsuperlattices by electron tomography. Nano Lett. 9, 2719–2724 (2009).

17. Talapin, D. V. et al. Quasicrystalline order in self-assembled binary nanoparticlesuperlattices. Nature 461, 964–967 (2009).

18. Jang, S., Kong, W. & Zeng, H. Magnetotransport in Fe3O4 nanoparticle arraysdominated by noncollinear surface spins. Phys. Rev. B 76, 212403 (2007).

19. Tauba, N., Tsukernikb, A. & Markovich, G. Inter-particle spin-polarized tunnelingin arrays of magnetite nanocrystals. J. Magn. Magn. Mater. 321, 1933–1938(2009).

20. Ramos, A. V. et al. Magnetotransport properties of Fe3O4 epitaxial thin films:thickness effects driven by antiphase boundaries. J. Appl. Phys. 100, 103902(2006).

21. Tran, T. B. et al. Multiple cotunneling in large quantum dot arrays. Phys. Rev. Lett.95, 076806 (2005).

22. Mueggenburg, K. E., Lin, X. M., Goldsmith, R. H. & Jaeger, H. M. Elasticmembranes of close-packed nanoparticle arrays. Nature Mater. 6, 656–660(2007).

23. Cheng, W. L. et al. Free-standing nanoparticle superlattice sheets controlled byDNA. Nature Mater. 8, 519–525 (2009).

24. Viau, G. et al. Ruthenium nanoparticles: size, shape, and self-assemblies. Chem.Mater. 15, 486–494 (2003).

25. Aleksandrovic, V. et al. Preparation and electrical properties of cobalt-platinumnanoparticle monolayers deposited by the Langmuir-Blodgett technique. ACSNano 2, 1123–1130 (2008).

26. Sachan, M., Walrath, N. D., Majetich, S. A., Krycka, K. & Kao, C. Interaction effectswithin Langmuir layers and three-dimensional arrays of e-Co nanoparticles. J.Appl. Phys. 99, 08C302 (2006).

27. Evers, W. H., Friedrich, H., Filion, L., Dijkstra, M. & Vanmaekelbergh, D.Observation of a ternary nanocrystal superlattice and its structuralcharacterization by electron tomography. Angew. Chem. Int. Ed. 48, 9655–9657(2009).

28. Park, J. et al. Ultra-large-scale syntheses of monodisperse nanocrystals. NatureMater. 3, 891–895 (2004).

29. Sun, S. H., Murray, C. B., Weller, D., Folks, L. & Moser, A. Monodisperse FePtnanoparticles and ferromagnetic FePt nanocrystal superlattices. Science 287,1989–1992 (2000).

30. Shevchenko, E. V. et al. Colloidal synthesis and self-assembly of CoPt3

nanocrystals. J. Am. Chem. Soc. 124, 11480–11485 (2002).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements A.D. and C.B.M. acknowledge the financial support from theUS Army Research Office (ARO) under award number MURI W911NF-08-1-0364.J.C., P.M.V. and J.M.K. are grateful for support from the NSF MRSEC programmeunder award number DMR-0520020. We thank C. Kagan for access to the thermalevaporator.

Author Contributions A.D. and C.B.M. conceived and designed the experiments.A.D. studied the Fe3O4–FePt and Fe3O4–CoPt3 nanocrystal systems, and J.C.studied the Fe3O4–Fe3O4 system and the CaB6-type Fe3O4–FePt system. A.D. andJ.C. carried out BNSL structural characterization and magnetoresistive devicefabrication. A.D., J.C., P.M.V. and J.M.K. studied the magnetotransport of BNSLmembranes. A.D. and C.B.M. wrote the paper. All authors discussed the results andcommented on the manuscript.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. The authors declare no competing financial interests.Readers are welcome to comment on the online version of this article atwww.nature.com/nature. Correspondence and requests for materials should beaddressed to A.D. (angang@sas.upenn.edu) or C.B.M.(cbmurray@sas.upenn.edu).

NATURE | Vol 466 | 22 July 2010 LETTERS

477Macmillan Publishers Limited. All rights reserved©2010