given their parallel spectral evolution (Fig. 2) we consider it morelikely that the two supernovae were coeval but displayed differentlight curves, as indicated by the differences in expansion velocity(Fig. 3b), peak brightness, detailed spectral evolution and strengthof associated GRB.

We now consider the properties of SN2003dh and SN1998bw inmore detail. Both belong to the unusual class of Type Ic supernovae.SN1998bw was well explained by models involving the very ener-getic explosion of a massive Wolf–Rayet star19,20 (a highly evolvedstar that has lost its outer hydrogen layers to a wind or companionstar). Since Type Ic supernovae have light curves powered byradioactive decay, the bright light curve indicates the synthesis ofat least several tenths of a solar mass of 56Ni in both events (ref. 21and the current paper). Modelled in spherical symmetry—whichis certainly a questionable hypothesis—SN1998bw required1–2 £ 1052 erg of kinetic energy19,20. The energy requirements hereare similar. However, it is much more likely that the explosion wasasymmetric; in fact the rapid rise and decay of SN2003dh seen inFig. 3 may indicate that we are viewing an asymmetric supernovaalong its axis of most rapid expansion9,20. The necessary energy maytherefore be considerably smaller20. Further, the larger ratio ofafterglow luminosity to supernova luminosity in GRB030329 isalso consistent with a burst that has been seen nearly pole-on22,whereas the low luminosity of GRB980425 and its afterglow mayhave been due to its having being viewed substantially off-axis, orsimply because it ejected less highly relativistic matter.

The combined results on SN1998bw and SN2003dh offer themost direct evidence yet that typical, long-duration, energetic GRBsresult from the deaths of massive stars. The lack of hydrogen lines inboth spectra is consistent with model expectations that the star lostits hydrogen envelope to become a Wolf–Rayet star before explod-ing. The broad lines are also suggestive of an asymmetric explosionviewed along the axis of most rapid expansion. The large 56Niabundance needed for the light curve cannot have been made by thelow-density jet; its mass and solid angle are too small. But it couldhave been produced in the wind from an accretion disk as matterflowed into a black hole9.

The temporal coincidence of SN2003dh with GRB030329 rulesout the supranova model23 for this event. Neither the jet itself, norany g-rays it made, could escape until the supernova had expandedfor several months. The presence of the large quantities of 56Ninecessary to account for the light curve may also be problematic inthe magnetar model24 where there is no accretion disk. Producingthe 56Ni in a shock would require that the pulsar’s energy bedistributed to a large mass, which is inconsistent with producinga relativistic jet with almost no mass. Our observations thus supportthe collapsar model8,9 for GRBs. A

Received 16 April; accepted 20 May 2003; doi:10.1038/nature01750.

1. Galama, T. J. et al. An unusual supernova in the error box of the g-ray burst of 25 April 1998. Nature

395, 670–672 (1998).

2. Paczynski, B. Are gamma-ray bursts in star-forming regions? Astrophys. J. 494, L45–L48 (1998).

3. Fruchter, A. S. et al. Hubble Space Telescope and Palomar imaging of GRB 990123: Implications for

the nature of gamma-ray bursts and their hosts. Astrophys. J. 519, L13–L16 (1999).

4. Bloom, J. S. et al. The unusual afterglow of the g-ray burst of 26 March 1998 as evidence for a

supernova connection. Nature 401, 453–456 (1999).

5. Castro-Tirado, A. J. & Gorosabel, J. Optical observations of GRB afterglows: GRB 970508 and GRB

980326 revisited. Astron. Astrophys. Suppl. Ser. 138, 449–450 (1999).

6. Garnavich, P. et al. Discovery of the low-redshift optical afterglow of GRB 011121 and its progenitor

supernova SN 2002ke. Astrophys. J. 582, 924–932 (2003).

7. Reeves, J. N. et al. The signature of supernova ejecta in the X-ray afterglow of the g-ray burst 011211.

Nature 416, 512–515 (2002).

8. Woosley, S. E. Gamma-ray bursts from stellar mass accretion disks around black holes. Astrophys. J.

405, 273–277 (1993).

9. MacFadyen, A. I. & Woosley, S. E. Collapsars: Gamma-ray bursts and explosions in ‘failed supernovae’.

Astrophys. J. 524, 262–289 (1999).

10. Kulkarni, S. R. et al. Radio emission from the unusual supernova 1998bw and its association with the

g-ray burst of 25 April 1998. Nature 395, 663–669 (1998).

11. Vanderspek, R. et al. GRB030329 (¼H2652): A long, extremely bright GRB localized by the HETE

WXM and SXC. GCN Circ. 1997 (2003).

12. Price, P. A. et al. The bright optical afterglow of the nearby g-ray burst of 29 March 2003. Nature 423,

844–847 (2003).

13. Torii, K. GRB 030329: OT candidate GCN Circ. 1986 (2003).

14. Greiner, J. et al. Redshift of GRB 030329. GCN Circ. 2020 (2003).

15. Patat, F. et al. The metamorphosis of SN 1998bw. Astrophys. J. 555, 900–917 (2001).

16. Garnavich, P., Matheson, T., Olszewski, E. W., Harding, P. & Stanek, K. Z. GRB 030329 and supernova

2003dh. IAU Circ. 8114 (2003).

17. Stanek, K. Z. et al. Spectroscopic discovery of the supernova 2003dh associated with GRB 030329.

Astrophys. J. Lett. (in the press); preprint at khttp://arXiv.org/astro-ph/0304173l (2003).

18. Leonard, D. C., Filippenko, A. V., Chornock, R. & Foley, R. J. Photospheric-phase spectropolarimetry

and nebular-phase spectroscopy of the peculiar Type Ic supernova 2002ap. Publ. Astron. Soc. Pacif.

114, 1333–1348 (2002).

19. Iwamoto, K. et al. A hypernova model for the supernova associated with the g-ray burst of 25 April

1998. Nature 395, 672–674 (1998).

20. Woosley, S. E., Eastman, R. G. & Schmidt, B. P. Gamma-ray bursts and Type Ic supernova SN 1998bw.

Astrophys. J. 516, 788–796 (1999).

21. Sollerman, J. et al. SN 1998bw at late phases. Astrophys. J. 537, L127–L130 (2000).

22. Granot, J., Panaitescu, A., Kumar, P. & Woosley, S. E. Off-axis afterglow emission from jetted gamma-

ray bursts. Astrophys. J. 570, L61–L64 (2002).

23. Vietri, M. & Stella, L. Supranova events from spun-up neutron stars: an explosion in search of an

observation. Astrophys. J. 527, L43–L46 (1999).

24. Wheeler, J., Yi, I., Hoflich, P. & Wang, L. Asymmetric supernovae, pulsars, magnetars, and gamma-ray

bursts. Astrophys. J. 537, 810–823 (2000).

25. Zharikov, S., Benitez, E., Torrealba, J. & Stepanian, J. GRB030329, SPM optical observations. GCN

Circ. 2022 (2003).

26. Schlegel, D. J., Finkbeiner, D. P. & Davis, M. Maps of dust infrared emission for use in estimation of

reddening and cosmic microwave background radiation foregrounds. Astrophys. J. 500, 525–553

(1998).

27. Kennicutt, R. C. Star formation in galaxies along the Hubble sequence. Annu. Rev. Astron. Astrophys.

36, 189–232 (1998).

28. Lee, H., Grebel, E. K. & Hodge, P. W. Nebular abundances of nearby southern dwarf galaxies. Astron.

Astrophys. 401, 141–159 (2003).

29. Blake, C. & Bloom, J. S. GRB 030329: Optical pre-imaging. GCN Circ. 2011 (2003).

30. Fynbo, J. U. et al. Hubble Space Telescope Space Telescope Imaging Spectrograph imaging of the host

galaxy of GRB 980425/SN 1998bw. Astrophys. J. 542, L89–L93 (2000).

Acknowledgements We thank F. Patat for discussions. This paper is based on observations

collected by the Gamma-Ray Burst Collaboration at ESO (GRACE) at the European Southern

Observatory, Paranal, Chile. We thank the ESO staff at the Paranal Observatory, in particular

N. Ageorges, P. Gandhi, S. Hubrig, R. Johnson, C. Ledoux, K. O’Brien, R. Scarpa, T. Szeifert and

L. Vanzi, for their help in securing the service mode data reported here. We acknowledge benefits

from collaboration within the EU FP5 Research Training Network “Gamma-Ray Bursts: An

Enigma and a Tool”. This work was also supported by the Danish Natural Science Research

Council (SNF). J.P.U.F. and K.P. acknowledge support from the Carlsberg Foundation.

Competing interests statement The authors declare that they have no competing financial

interests.

Correspondence and requests for materials should be addressed to J.H. (jens@astro.ku.dk).

..............................................................

Beating the superparamagnetic limitwith exchange biasVassil Skumryev*, Stoyan Stoyanov*, Yong Zhang*,George Hadjipanayis*, Dominique Givord† & Josep Nogues‡

* Department of Physics and Astronomy, University of Delaware, Newark,Delaware 19716, USA† Laboratoire Louis Neel, CNRS, 38042 Grenoble cedex 9, France‡ Institucio Catalana de Recerca i Estudis Avancats (ICREA) and Department deFısica, Universitat Autonoma de Barcelona, 08193 Bellaterra, Spain.............................................................................................................................................................................

Interest in magnetic nanoparticles has increased in the past fewyears by virtue of their potential for applications in fields such asultrahigh-density recording and medicine1–4. Most applicationsrely on the magnetic order of the nanoparticles being stablewith time. However, with decreasing particle size the magneticanisotropy energy per particle responsible for holding themagnetic moment along certain directions becomes comparableto the thermal energy. When this happens, the thermal fluctua-tions induce random flipping of the magnetic moment with time,and the nanoparticles lose their stable magnetic order andbecome superparamagnetic5. Thus, the demand for further

letters to nature

NATURE | VOL 423 | 19 JUNE 2003 | www.nature.com/nature850 © 2003 Nature Publishing Group

miniaturization comes into conflict with the superparamagnet-ism caused by the reduction of the anisotropy energy per particle:this constitutes the so-called ‘superparamagnetic limit’6,7 inrecording media. Here we show that magnetic exchange couplinginduced at the interface between ferromagnetic and antiferro-magnetic systems8,9 can provide an extra source of anisotropy,leading to magnetization stability. We demonstrate this principlefor ferromagnetic cobalt nanoparticles of about 4 nm in diameterthat are embedded in either a paramagnetic or an antiferromag-netic matrix. Whereas the cobalt cores lose their magneticmoment at 10 K in the first system, they remain ferromagneticup to about 290 K in the second. This behaviour is ascribed to thespecific way ferromagnetic nanoparticles couple to an antiferro-magnetic matrix.

The system studied consists of Co nanoparticles embedded eitherin a paramagnetic matrix (C or Al2O3) or in an antiferromagnetic(AFM) matrix (CoO). The samples are grown by sequentiallydepositing a layer of matrix material of thickness 15–20 nm followedby a layer of nanoparticles, up to a total of 10 repeats (Fig. 1). The Conanoparticles were produced by gas condensation of sputteredatoms in a so-called ‘cluster gun’10, incorporated inside a conven-tional sputtering apparatus. Some oxygen (7:1 Ar:O2 ratio) wasallowed inside the sputtering chamber to oxidize the surface of thenanoparticles, forming a CocoreCoOshell core–shell structure. Thispartial oxidation produces a passivating layer11, and, in the sampleswith an AFM CoO matrix, ensures the formation of a continuousCo-core/CoO-matrix interface. The yield of the cluster gun wascontrolled to obtain a rather disperse spatial distribution of nano-particles. The matrix material was deposited either by conventionald.c. sputtering (C) or by reactive sputtering with an O2 partialpressure of 0.7 mtorr (CoO and Al2O3). As revealed by transmissionelectron microscopy (TEM), the Co particles have a roughlyspherical core of d core < 3–4 nm in diameter and a face-centred

cubic (f.c.c.) structure (Fig. 1). The shell is f.c.c. CoO with athickness of about t shell < 1 nm. The CocoreCoOshell nanoparticlesexhibit a log-normal distribution with a mean dcore þ t shell < 4.7nm and a standard deviation of 1.1 nm. The average in-plane andout-of plane interparticle distances are around 12 nm, significantlylarger than the diameter of the nanoparticles. As the maximumdipolar field, m0H dip (where m0 is the permeability of vacuum),between two nanoparticles in contact amounts to 0.07 T, in noinstance may interparticle dipolar interactions be the source of theeffects described below.

The temperature (T) dependence of the magnetic moment m(T)under an applied magnetic field (m0Happ ¼ 0.01 T) was measuredafter field cooling (FC) and after zero-field cooling (ZFC). Theblocking temperature, TB, is the temperature at which superpar-amagnetism sets in. Below TB, the FC and ZFC magnetizationcurves are split, whereas above TB, they coincide as the remanenceand coercivity have vanished. For 4-nm particles embedded in anAl2O3 (paramagnetic) matrix, T B < 10 K (Fig. 2). Similar resultswere obtained for a C matrix. Increasing the particle size to about7 nm resulted in an enhancement of TB up to 200 K, in reasonableagreement with published results12–19 (which are rather spread dueto structural differences affecting the anisotropy K U, and the effectof dipolar interactions). The hysteresis loops of both systems (withAl2O3 and C matrices) measured below T B are characterized by asmall value of coercivity (m0H C ¼ 0.02 T) (Fig. 3a) and no loopshift on the magnetic field axis for samples field cooled in fields upto 5 T. Exchange bias9 at ferromagnetic (FM)/AFM interfaces ischaracterized by (1) coercivity enhancement, revealing induceduniaxial or multiaxial anisotropy, and (2) hysteresis loop shiftalong the field axis after field cooling, revealing unidirectionalanisotropy. The lack of exchange bias in the non-magnetic matrixsystems can tentatively be attributed to the small thickness of theAFM CoO shell11,20. In any case, the exact interpretation of thisbehaviour will not affect the discussion presented below.

When such CocoreCoOshell particles were deposited withoutdilution in a paramagnetic matrix and hence in close contact witheach other, they were found to exhibit all the features of an exchangebiased system (Fig. 3b)—exchange bias field of about moH eb < 0.92T and enhanced coercivity (moH C ¼ 0.39 T in the ZFC case andmoH C ¼ 0.59 T in the FC one) at 4.2 K. The bias field decreased as

Figure 1 TEM micrographs and electron diffraction of CocoreCoOshell particles. a, High-

magnification bright-field image, revealing the core–shell structure. Inset, plan-view

distribution of the nanoparticles. b, High-resolution lattice image of nanoparticle with

[001]f.c.c. crystallographic orientation. c, Electron-diffraction patterns, showing f.c.c. Co

and f.c.c. CoO reflections. d, Schematic drawing of the sample cross-section, showing Co

cores (black), and surrounding CoO shell (white) and matrix (grey).

Figure 2 Magnetic moments of 4-nm CocoreCoOshell particles. Shown is the temperature

dependence of the zero-field cooled (ZFC; filled symbols) and field-cooled (FC;

m 0H FC ¼ 0.01 T, open symbols) magnetic moment (m) of 4-nm CocoreCoOshell particles.

Particles were embedded in a paramagnetic (Al2O3) matrix (diamonds), or in an AFM (CoO)

matrix (circles). The measuring field is m 0H ¼ 0.01 T. The Neel temperature of CoO is

indicated by an arrow. The lines are guides to the eye.

letters to nature

NATURE | VOL 423 | 19 JUNE 2003 | www.nature.com/nature 851© 2003 Nature Publishing Group

temperature increased, and vanished at about 180 K, in agreementwith the classical data for CocoreCoOshell compacted particles8,11,19,21.This difference in the behaviour of isolated and compactedCocoreCoOshell particles emphasizes the important role of interpar-ticle coupling in stabilizing not only the ferromagnetism of theparticle core but also the antiferromagnetism of the particle shell.Note that this stabilization of FM nanoparticle magnetizationthrough interparticle magnetic interactions is not of interest formagnetic recording.

The magnetic behaviour of the isolated CocoreCoOshell particleschanges markedly when, instead of being embedded in a paramag-netic matrix, they are embedded in an AFM CoO matrix of similarthickness. As can be seen from the ZFC-FC m(T) curve (Fig. 2), theCo nanocores remain FM up to the Neel temperature, T N, of CoO(TN < 290 K). This indicates that an extra anisotropy is inducedsuch that K UV .. k BT, where V is the particle volume and k B isBoltzmann’s constant. In this case, the nanocore moments areprevented from flipping over the energy barrier for all temperaturesbelow TN of CoO, and thus the nanoparticles remain magneticallystable below this temperature. A hysteresis loop typical of Cocore-

CoOshell nanoparticles embedded in a CoO matrix is shown inFig. 3c. Below TN it is characteristic of an exchange biased system,exhibiting an exchange bias field of moHeb ¼ 0.74 T and enhancedcoercivity of moH C ¼ 0.76 T at 4.2 K. The magnetic stability of thenanoparticles is further evidenced in Fig. 4, where both H C and theremanent moment, mR remain much larger than zero for T , T N.In contrast, the Co nanoparticles embedded in a paramagneticmatrix have zero mR and H C for T . TB ¼ 10 K. A behavioursimilar to the one described above for CocoreCoOshell particles was

also observed for pure Co nanoparticles (that is, without CoO shell)embedded in a CoO matrix. However, because of the poorer qualityof the interface between the AFM matrix and the Co FM nanopar-ticles, all the effects associated with exchange anisotropy were lesspronounced. Note that additional measurements on diamagneticAg nanoparticles embedded in a CoO matrix did not show FMresponse, thus confirming that the above effects originate from theFM Co nanocores and not from a possible weak ferromagnetismdue to the CoO matrix. Altogether, these results demonstrate thatthe coupling of FM particles with an AFM matrix can be a source ofa large effective anisotropy, leading to a considerable increase of thenanoparticle blocking temperature. We note that the recent obser-vation of loop squareness enhancement in FM particles embeddedin or grown on an AFM film22,23 has already indicated improvedmagnetic stability. Moreover, exchange bias has already been pro-posed theoretically as a possible source of magnetization stability24.However, such large increases in TB, as well as coercive fields andexchange bias fields of the order of those required to account for thepresent data, have not been previously observed.

FM-AFM exchange coupling may be understood by assumingthat the FM moments of the Co particles create an exchange field,m0H ex, acting on the interface AFM moments of the CoO matrix.CoO is a collinear antiferromagnet. At a compensated interface, thenumber of AFM moments from each sublattice is the same, theresulting magnetization is zero and thus the coupling vanishes tofirst order. To second order, the exchange field induces a canting ofthe AFM moments, and a FM component is induced along the FMCo moments25,26. Considering that Hex is formally equivalent to anexternal field, the coupling energy is5 E ¼ 21/2 xAFm0H2

ex; wherexAF is the interface AFM susceptibility. The maximum in xAF

determines the FM ‘easy’ direction. The energy difference betweentwo FM moment directions, DE, amounts to 2(1/2) DxAFm0H2

ex;where DxAF is the associated difference in susceptibility. DE may beviewed as an additional anisotropy term, acting as a source ofmagnetization stability of the Co nanoparticle at finite temperature.Moreover, under an applied field, antiparallel to the FM magnetiza-tion, it constitutes a barrier against magnetization reversal. At suchvery small nanoparticle sizes, the formation of heterogeneousmagnetization state costs energy. We found that DE is minimumfor coherent rotation of the FM moments. The possibility of FMmoment rotation within the AFM easy plane, as well as within theplane P perpendicular to the AFM moment direction, were exam-ined. We found that DE is minimum for moment rotation withinP. The deduced coercive field is m0Hc ¼ DEmin=MSVFM; where

Figure 3 Hysteresis loops at 4.2 K of 4-nm CocoreCoOshell particles embedded in different

matrices. Data are shown after ZFC (dashed lines) and FC (m 0H FC ¼ 5 T; solid lines)

procedures. a, Embedded in a paramagnetic (Al2O3) matrix; b, compacted; and

c, embedded in an AFM (CoO) matrix.

Figure 4 Coercivity and remanence of 4-nm CocoreCoOshell particles. Shown is the

temperature dependence of coercivity, H C (filled symbols), and remanence, m R (open

symbols), after ZFC, of 4-nm CocoreCoOshell particles embedded in different matrices.

Squares, a paramagnetic (Al2O3) matrix; circles, an AFM (CoO) matrix. The lines are

guides to the eye.

letters to nature

NATURE | VOL 423 | 19 JUNE 2003 | www.nature.com/nature852 © 2003 Nature Publishing Group

DE min represents the DE minimum, VFM is the FM nanoparticlevolume and MS is the saturation magnetization. m0H C ¼ 1.3 T iscalculated for the parameter values of AFM sublattice magnetiza-tion5 m 0M 0

¼ 1.8 T, a molecular field5 coefficient between theAFM sublattices of w ¼ 100, an AFM second-order anisotropyconstant of K 1 ¼ 2.7 £ 107 J m23 (ref. 27), V FM ¼ 33.5 nm3 andmoH ex ¼ 84 T (ref. 28). Considering the uncertainty in the values ofmost of these parameters, the agreement with the experimentalvalue, m 0H C ¼ 0.76 T, is very satisfactory. Unlike the usualexchange-bias systems, the calculated and experimental values ofthe coercive field are in agreement. Owing to their very small size,the FM particles can be thought of as being coupled to a uniqueAF domain, and consequently there is no competition betweendifferent coupling terms.

Additionally, exchange coupling at the Co/CoO interface givesrise to a shift of the hysteresis loop along the field axis. The large biasfields observed may be shown to suggest the existence of AFM non-collinear moment configurations, reminiscent of spin-glass con-figurations. Such configurations exist in magnetic oxide nanopar-ticles29,30, and, by analogy, they may be expected at the surface ofsmall cavities.

In all the above discussion, a compensated Co/CoO interface hasbeen assumed. Coupling energies of the same order of magnitudecould be obtained in the case of large uncompensation, of the orderof 30%, existing at the Co/CoO interface. However, in this case asignificant part of the FM signal would originate from the uncom-pensated moments in the matrix. The measurements on the Ag/CoO sample do not reveal such signal, and large uncompensationcan thus be realistically excluded.

The present results demonstrate that the magnetic coupling ofFM nanoparticles with an AFM matrix is a source of a large effectiveadditional anisotropy. This leads to a marked improvement in thethermal stability of the moments of the FM nanoparticles—weobserved an increase in the blocking temperature, TB, of almosttwo orders of magnitude. This mechanism provides a way to beatthe ‘superparamagnetic limit’ in isolated particles. Although it isclear that the system examined here is not suitable in itself forapplication, the approach developed should in principle apply tonanoparticles deposited on a single AFM layer, a structure suitablefor use as a recording medium. With the right choice of FM andAFM components, exchange anisotropy coupling could ultimatelyallow magnetically stable dots only a few nanometres in size: suchdots would surpass the storage-density goal of 1 Tbit in22, as set bythe magnetic-storage industry. A

Received 24 December 2002; accepted 22 April 2003; doi:10.1038/nature01687.

1. Kodama, R. H. Magnetic nanoparticles. J. Magn. Magn. Mater. 200, 359–372 (1999).

2. Martın, J. I. et al. Ordered magnetic nanostructures: Fabrication and properties. J. Magn. Magn.

Mater. 256, 449–501 (2003).

3. Sun, S. H. et al. Monodisperse FePt nanoparticles and ferromagnetic FePt nanocrystal superlattices.

Science 287, 1989–1992 (2000).

4. Hafeli, U., Schutt, W., Teller, J. & Zborowski, M. (eds) Scientific and Clinical Applications of Magnetic

Materials (Plenum, New York, 1997).

5. Chikazumi, S. Physics of Ferromagnetism (Oxford Univ. Press, New York, 1997).

6. Weller, D. & Moser, A. Thermal effect limits in ultrahigh-density magnetic recording. IEEE Trans.

Magn. 35, 4423–4439 (1999).

7. Thompson, D. A. & Best, J. S. The future of data storage technology. IBM J. Res. Dev. 44, 311–322

(2000).

8. Meiklejohn, W. H. & Bean, C. P. New magnetic anisotropy. Phys. Rev. 102, 1413–1414 (1956).

9. Nogues, J. & Schuller, I. K. Exchange bias. J. Magn. Magn. Mater. 192, 203–232 (1999).

10. Huberland, H. et al. Thin film growth from energetic cluster impact: A feasibility study. J. Vac. Sci.

Technol. A 10, 3266–3271 (1992).

11. Gangopadhyay, S. et al. Exchange anisotropy in oxide passivated Co fine particles. J. Appl. Phys. 73,

6964–6966 (1993).

12. Luis, F. et al. Magnetic relaxation of interacting Co clusters: Crossover from two- to three-dimensional

lattices. Phys. Rev. Lett. 88, 217205 (2002).

13. Garcıa-Otero, J. et al. Influence of dipolar interaction on magnetic properties of ultrafine

ferromagnetic particles. Phys. Rev. Lett. 84, 167–170 (2000).

14. Gu, E. et al. Two-dimensional paramagnetic-ferromagnetic phase transition and magnetic anisotropy

in Co(110) epitaxial nanoparticle arrays. Phys. Rev. B 60, 4092–4095 (1999).

15. Parent, F. et al. Giant magnetoresistance in Co-Ag granular films prepared by low-energy cluster beam

deposition. Phys. Rev. B 55, 3683–3687 (1997).

16. Held, G. A. et al. Competing interactions in dispersions of superparamagnetic nanoparticles. Phys.

Rev. B 64, 012408 (2001).

17. Woods, S. I. et al. Direct investigation of superparamagnetism in Co nanoparticle films. Phys. Rev. Lett.

87, 137205 (2001).

18. Murray, C. B. et al. Monodisperse 3d transition-metal (Co,Ni,Fe) nanoparticles and their assembly

into nanoparticle superlattices. Mater. Res. Soc. Bull. 26, 985–991 (2001).

19. Sako, S., Osshima, K. & Saki, M. Magnetic property of oxide passivated Co nanosized particles

dispersed in two dimensional plane. J. Phys. Soc. Jpn 70, 2134–2138 (2001).

20. Lund, M. S. et al. Effect of anisotropy on the critical antiferromagnetic thickness in exchange biased

bilayers. Phys. Rev. B 66, 054422 (2002).

21. Peng, D. L. et al. Magnetic properties of monodispersed Co/CoO clusters. Phys. Rev. B 61, 3103–3109

(2000).

22. Sort, J. et al. Room temperature coercivity enhancement in mechanically alloyed antiferromagnetic-

ferromagnetic powders. Appl. Phys. Lett. 75, 3177–3179 (1999).

23. Liu, K. et al. Fabrication and thermal stability of arrays of Fe nanodots. Appl. Phys. Lett. 81, 4434–4436

(2002).

24. Jensen, P. J. Magnetic recording medium with improved temporal stability. Appl. Phys. Lett. 78,

2190–2192 (2001).

25. Schulthess, T. C. & Butler, W. H. Consequences of spin-flop coupling in exchange biased films. Phys.

Rev. Lett. 81, 4516–4519 (1998).

26. Stamps, R. L. Mechanisms for exchange bias. J. Phys. D 33, R247–R268 (2000).

27. Kanamori, J. Theory of the magnetic properties of ferrous and cobaltous oxides. Prog. Theor. Phys. 17,

177–196 (1957).

28. Takano, K. et al. Interfacial uncompensated antiferromagnetic spins: Role of unidirectional

anisotropy in polycrystalline Ni81Fe19/CoO bilayers. Phys. Rev. Lett. 79, 1130–1133 (1997).

29. Kodama, R. H., Makhlouf, S. H. & Berkowitz, A. E. Finite size effects in antiferromagnetic NiO

nanoparticles. Phys. Rev. Lett. 79, 1393–1396 (1997).

30. Coey, J. M. D. Noncollinear spin arrangement in ultrafine ferrimagnetic crystallites. Phys. Rev. Lett. 27,

1140–1143 (1971).

Acknowledgements We thank N. Dempsey for critical reading of the manuscript, and D. Weller

for discussions. This work was partly supported by the US National Science Foundation, Seagate

Technology, the Catalan Direccio General de Recerca, and the Spanish Comision Interministerial

de Ciencia y Tecnologıa.

Competing interests statement The authors declare that they have no competing financial

interests.

Correspondence and requests for materials should be addressed to V.S. (vassil@udel.edu).

..............................................................

Sea-level fluctuations during the lastglacial cycleM. Siddall*, E. J. Rohling*, A. Almogi-Labin†, Ch. Hemleben‡,D. Meischner§, I. Schmelzer‡ & D. A. Smeed*

* Southampton Oceanography Centre, European Way, Southampton SO14 3ZH,UK† Geological Survey of Israel, 30 Malkhe Yisrael Street, Jerusalem 95501, Israel‡ Department of Geology and Paleontology, University of Tuebingen,Sigwartstrasse 10, D-7400, Tuebingen, Germany§ Department of Sedimentary Geology, Institute for Geology and Paleontology,University of Gottingen, 37073 Gottingen, Germany.............................................................................................................................................................................

The last glacial cycle was characterized by substantial millennial-scale climate fluctuations1–5, but the extent of any associatedchanges in global sea level (or, equivalently, ice volume) remainselusive. Highstands of sea level can be reconstructed from datedfossil coral reef terraces6,7, and these data are complemented by acompilation of global sea-level estimates based on deep-seaoxygen isotope ratios at millennial-scale resolution8 or higher1.Records based on oxygen isotopes, however, contain uncertain-ties in the range of 630 m, or 61 8C in deep sea temperature9,10.Here we analyse oxygen isotope records from Red Sea sedimentcores to reconstruct the history of water residence times in theRed Sea. We then use a hydraulic model of the water exchangebetween the Red Sea and the world ocean to derive the sill depth—and hence global sea level—over the past 470,000 years (470 kyr).Our reconstruction is accurate to within 612 m, and gives acentennial-scale resolution from 70 to 25 kyr before present. We

letters to nature

NATURE | VOL 423 | 19 JUNE 2003 | www.nature.com/nature 853© 2003 Nature Publishing Group