Weak Ferromagnetism in Fe Co Sb 1−x x 2 Rongwei Hu1,2, R. P. Hermann3∗, F. Grandjean3, Y. Lee4, J. B. Warren5, V. F. Mitrovi´c2 and C. Petrovic1 1Condensed Matter Physics, Brookhaven National Laboratory, Upton New York 11973-5000 USA 2Physics Department, Brown University, Providence RI 02912 3Department of Physics B5, Universit´e de Li`ege, Belgium 4Department of Earth System Sciences, Yonsei University, Seoul 120749, Korea 5Instrumentation Division, Brookhaven National Laboratory, Upton New York 11973-5000 USA (Dated: February 2, 2008) WeakferromagnetisminFe1−xCoxSb2isstudiedbymagnetizationandM¨ossbauermeasurements. 8 Asmallspontaneousmagneticmomentoftheorderof∼10−3µB appearsalongthebb-axisfor0.2≤ 0 x≤0.4. Basedonthestructuralanalysis,weargueagainstextrinsicsourcesofweakferromagnetism. 0 We discuss our results in the framework of the nearly magnetic electronic structure of the parent 2 compound FeSb2. n PACSnumbers: 75.30.-m,76.80.+y,71.28.+d a J 1 I. INTRODUCTION InanalogytoFeSi,recentab-initiocalculationspredicted 1 the nearly ferromagnetic nature of the FeSb ground 2 state.18 InFeSitheferromagneticstatehasbeeninduced el] froFmeSaiannodnmFeaSgbn2etaircesseemmicicoonndduuccttinorgsgtrhoautnsdhoswtacterowssiothvear by lattice expansion in FeSi1−xGex19 or by carrier in- - narrow gap to a thermally induced paramagnetic metal sertion in Fe1−xCoxSi.20 In contrast, FeSb2 has not yet tr with enhanced susceptibility.1,2 The magnetic proper- been tuned to a ferromagnetic state by any external pa- s rameters. In this work, we demonstrate the presence tiesofFeSihaveinstigatedconsiderabletheoreticalinter- . at est, starting with the narrow-band model of Jaccarino.3 of the weak ferromagnetism (WFM) in Fe1−xCoxSb2 (0.2≤x≤0.45). Theoriginsofthe WFMarediscussed. m Further models include a nearly ferromagnetic semicon- ductor model of Takahashi and Moriya4 in which the Extensive structural analysis shows no evidence of ex- - d state was sustained by thermally induced spin fluctua- trinsic impurity induced WFM. We argue that instead n tions found in neutron scattering experiments.5,6 More- the WFM is a consequence of the nearly ferromagnetic o electronic structure of the parent compound FeSb . over,thenearlyferromagneticsemiconductorpicturewas 2 c supported by LDA+U band structure calculations by [ Mattheiss and Hamann7 and Anisimov et al.8 At the 1 same time, Aeppli and Fisk9 pointed out that the mag- II. EXPERIMENT v netic properties of FeSi are analogous to the physics of 8 8 Kondoinsulators,albeitwithareducedon-siteCoulomb The Fe1−xCoxSb2 single crystals were grown from ex- 6 repulsion U. The basis of their argument was a model, cess Sb flux.2 Powder X-ray diffraction (XRD) patterns 1 ruledoutbyJaccarinoinhisoriginalwork,ofthenarrow of the ground samples were taken with Cu K radi- α . gap and high density of states. Experiments of Mandrus ation (λ = 1.5418 ˚A) using a Rigaku Miniflex X-ray 1 0 et al.10 and Park et al.11 confirmed the validity of the diffractometer. The lattice parameters were obtained 8 model of Aeppli and Fisk. using Rietica software.21 High resolution XRD patterns 0 A search for new model systems, where the applica- were taken at the beamline X7A of the National Syn- : bility of the Kondo insulator framework to 3d transition chrotron Light Source at the Brookhaven National Lab- v i metals can be investigated, led to the synthesis of large oratoryusingmonochromaticsynchrotronX-rayandgas- X single crystals of FeSb2. Furthermore, a crossover was proportional position-sensitive detector. Rietveld refine- r discoveredsimilartotheoneinFeSi,forthemagneticand mentswereperformedusingGSAS.22 AJEOLJSM-6500 a electrical transport properties.2,12 Subsequent alloying SEM microprobe with resolution of 1.5 nm was used for studies haveshownheavyfermionmetallic state induced verifyingtheCoconcentrationsandinvestigatingthemi- in FeSb2−xSnx, just as in FeSi1−xAlx.13,14 In both ma- crostructure. Single crystals were oriented using a Laue terials the optical conductivity revealed unconventional Camera. Magnetization measurements were performed charge gap formation. That is, a complete recovery of in a Quantum Design MPMS XL 5 instrument. The spectral weight in FeSi and FeSb2 occurs over an energy iron-57Mo¨ssbauerspectra,attemperaturesrangingfrom rangeoffeweV,suggestingcontributionsoflargerenergy 2.8 to 295 K, were measured on a constant acceleration scales.15,16 This is in sharp contrast to metal-insulator spectrometer that utilized a rhodium matrix cobalt-57 transitions in band insulators where thermal excitations source. The instrument was calibrated at 295 K with ofchargecarriersthroughthegapredistributejustabove α-ironpowder. Theisomershiftsreportedhereinarerel- the gap. ative to α-iron at 295 K. The thickness of the absorber One of the key predictions of the LDA+U approach was23and72mg/cm2forFeSb andFe Co Sb ,re- 2 0.75 0.25 2 wasthecloseproximityofFeSitoaferromagneticstate.17 spectively. The sample temperature in the Janis SV-300 2 cryostat was controlled with a LakeShore 330 tempera- Fig. 2. For field strength varying between -6 and 6 kOe b turecontrollerandasilicondiodemountedonthecopper applied along the b - axis, hysteresis loops are observed sample holder. The accuracy of the sample temperature for 0.20 ≤ x ≤ 0.45. The width of the hysteresis loop is better than ±1%. growsinitially with increasing x from x=0.20,peaks at The powder X-ray patterns show that the x=0.25,andbecomesprogressivelysmalleruponfurther Fe1−xCoxSb2 (0.2 ≤ x ≤ 0.45) samples crystallize Co substitution. Hysteresis loops are absentfor field ap- in the Pnnm structure without any additional crys- pliedalongthebc -axisandareobservedonlyforx=0.25 talline peaks introduced by Co alloying. The effect of for field applied along the ab - axis. By extrapolating the Co substitution on the Fe site is to expand the unit magnetization of Fe Co Sb to H=0, a lower esti- 0.75 0.25 2 cell volume as compared to FeSb2. This expansion mate ofthe saturationmagnetizationalong the b-axis of is anisotropic and results from a contraction in the M = 0.0005 µ /F.U.) or (5.10−4 µ /Fe), where F.U. LL B B basal a-b plane and an expansion along the c-axis upon refers to the FeSb formula unit, is obtained. 2 substitution of Fe by Co.23 III. MAGNETIC PROPERTIES 2.0 2 U.11..05 -3M (10B)/F.U.-011 H II a xx==00..2205 B)/F.0.5 -2-6 -4 -2H(k0Oe)2 4 6 xxx===000...344005 5 5 H II a -3 00.0 2 -3 H (10 emu/mole)234 -3 M/H (10 emu/mole)1234 HH IIII bc M (1----2110....0505-6 H II -b4 -2 0 -3M (10B)/F.U.--21012-6 -4 -2H(0k4Oe)H2 II c4 66 M/ 0 H (kOe) 0 1 2 3 4 5 6 7 8 9 10 1 T(K) FIG. 2: Hysteresis loops for Fe1−xCoxSb2 in the ferromag- neticstate(x=0.2−0.45) at T=1.8K.Magnetization does 0 not saturate, it continues to increase with applied magnetic 0 50 100 150 200 250 300 350 field, similar to bulkitinerant ferromagnets with 3d ions. T(K) FIG.1: MagneticsusceptibilityM/HofFeSb2(opensymbols) The Mo¨ssbauer spectra of FeSb single crystals ex- 2 and Fe0.75Co0.25Sb2 (filled symbols) for a 1kOe field applied hibit a doublet at T=295 and 4.2 K. No impurity, and along all three principal crystalline axes. in particular no impurity with a large hyperfine field, is observed in the Mo¨ssbauer spectra. Furthermore, At low temperature, the parent compound FeSb is a the Mo¨ssbauer spectral parameters for FeSb obtained 2 2 narrow gap semiconductor with a rather small and tem- herein are in excellent agreementwith the previously re- peratureindependentmagneticsusceptibility.2Similarto ported parameters (see Table I).26 For Fe Co Sb , 0.75 0.25 2 FeSi, above 100 K there is a temperature induced para- the Mo¨ssbauer spectra, shown in Fig. 3, exhibit a dou- magnetic susceptibility and an enhanced electronic con- blet for temperatures ranging from T =295 K to 2.8 K. duction. Themagneticsusceptibilitycanbedescribedby Again no impurity contribution is observed. Its spectral bothathermallyinducedPaulisusceptibilityandalowto parameters,obtainedatT =295K and4K,arecloseto highspin transition.2,12,25 Inthe temperature (T)range thoseobservedintheFeSb . Theisomershiftobservedin 2 from 1.7 to 150 K the Fe Co Sb magnetic suscep- Fe Co Sb is ca. 0.01 mm/s smaller than in FeSb . 0.75 0.25 2 0.75 0.25 2 2 tibility is larger than that of FeSb . For T above 6 K, This indicates a somewhat larger s-electron density at 2 it shows little anisotropywith the magnetic field applied the 57Fe nucleus. along the different crystallographic axes. As shown in The variation of the quadrupole splitting from 295 to Fig. 1, the temperature dependence of the susceptibility 4.2 K is larger in FeSb than in Fe Co Sb . This 2 0.75 0.25 2 indicates Pauli paramagnetism at high temperature. A strong temperature dependence of the quadrupole split- clearferromagnetictransitionatT =6Kforafieldof1 tinginFeSb isconsistentwithascenarioofelectronde- C 2 kOe applied along any of the three crystallographicaxes localizationappearingwithincreasingtemperature,with is illustrated in the inset to Fig. 1. These observations a gap ∆E of 380 K.25,27 As illustrated in Fig. 4, a fit are in agreement with ferromagnetic long range order of of the Fe Co Sb quadrupole splitting as a func- 0.75 0.25 2 the small magnetic moments below T = 5 K.23 The fer- tion of temperature with the delocalization model de- romagnetic nature of the transition is supported by the scribed in Ref. 24 yields a somewhat larger gap energy hysteresis loop measured at T = 1.8 K and displayed in E =(480±50)KthanthatobservedinFeSb . Thedif- g 2 3 TABLE I: The hyperfine parameters for FeSb2 and Fe0.75Co0.25Sb2a; aRelative to α-iron at 295K and bThe isomer shift reference in Ref. 24 is sodium nitroprusside, which hasa -0.26 mm/s isomer shift relative to α-iron at room temperature. Fe0.75Co0.25Sb2 FeSb2 T(K) δa,mm/s ∆EQ,mm/s Γ,mm/s δa,mm/s ∆EQ,mm/s Γ,mm/s 296b - - - 0.450(6) 1.286(6) - 295 0.433(2) 1.343(2) 0.264(5) 0.449(1) 1.275(2) 0.262(3) 240 0.483(2) 1.394(2) 0.265(2) - - - 190 0.509(2) 1.422(2) 0.267(2) - - - 140 0.538(2) 1.449(2) 0.273(2) - - - 90 0.455(2) 1.364(2) 0.284(2) - - - 50 0.569(2) 1.474(3) 0.290(4) - - - 6.4b - - - 0.572(6) 1.575(6) - 4.2 0.560(2) 1.483(2) 0.291(2) 0.572(1) 1.573(3) 0.270(4) 2.8 0.558(2) 1.483(3) 0.338(4) - - - FIG. 4: Fit of the Fe0.75Co0.25Sb2 quadrupole splitting as a function of temperature with thedelocalization model. ent. Furthermore no iron-bearing impurity is observed in Fe Co Sb . 0.75 0.25 2 Apparently,theT=2.8KspectrumofFe Co Sb 0.75 0.25 2 is a doublet, which is somewhat surprising. Our inter- pretationis that either the ironexperiences no magnetic FIG.3: The57FeM¨ossbauerspectraofFe0.75Co0.25Sb2atthe hyperfine field or that the hyperfine field is below the indicated temperatures. The solid line is a fit to a doublet, detection limit. If the small broadening of ca. 0.047(6) using theparameters indicated in Table 1. mm/s of the 2.8 K spectrum, when compared to the 4.2 Kspectrum,wasassociatedtoamagnetichyperfinefield, it would correspond to a 1.5+/-0.2 kOe hyperfine field. ference between the hyperfine parameters in FeSb and With a linewidth constrained to 0.29 mm/s, a fit of this 2 Fe Co Sb indicates that there is indeed a modifi- spectrum, with both a quadrupole interaction and a hy- 0.75 0.25 2 cation of the FeSb structure, and that no phase segre- perfine field yields a field of 2.8+/-1.2 kOe. Taking the 2 gation is present. The Mossbauer spectra show that the usual proportionality of ca. 150 kOe/µ , these values B investigated phase is (Fe,Co)Sb and not FeSb +CoSb can be used to estimate an upper limit of about M = 2 2 2 UL since the hyperfine parameters are significantly differ- 0.01 µ . for the magnetic moment on Fe. B 4 200 Synchrotron XRD FeSb2 structure (calculated) 180 -Fe structure (calculated) 2 Fe0.7Co0.3Sb2 160 1.8K 10K 1 140 U.) 100K y (341) /F.B 0 300K ntensit120 (151) -3 10 aled i100 M( Sc -1 80 60 (042) -2 40 (250) -6 -4 -2 0 2 4 6 (322) Fe (112) 20 H(kOe) 0 FIG. 5: Magnetic hysteresis for theFe0.7Co0.3Sb2 sample. It 35.0 35.5 36.0 36.5 is important to note the absence of hysteresis loops above 2 (deg.) the ferromagnetic transition of FeSb2. Data were taken on a FIG. 6: Observed (black) and calculated (red) synchrotron crystal from independentlygrown batch. powderX-raydiffractionpatternsofFe0.75Co0.25Sb2. Calcu- latedpatternincludes0.3%ofsuperimposedα-Feimpurity. If present, impurity would have caused detectable deviation IV. INTRINSIC VS. EXTRINSIC MAGNETISM of the observed pattern since there is no peak overlap. Given the small value of the saturated moment, it is possible that WFM originates from extrinsic sources, suchasartifactsofthe measurementprocessorthe pres- ence of a small amount of ferromagnetic impurity, e.g. elemental Fe. The former can be excluded based on the lack of sample dependence, both in magnetization and in heat capacity data.23 Below we discuss the possibility of undetected second phases as extrinsic sources of the WFM. No hysteresis loops are observed for temperatures aboveT =(6−7)K forx=(0.2−0.4)(example shown C in Fig. 5). No known Fe-Sb, Co-Sb, Fe-Co, Fe-O, or Co- FIG. 7: Typical Scanning Electron Microscope (SEM) image Ophasesshowaferromagnetictransitioninthistemper- of Fe0.75Co0.25Sb2. SEM images on randomly chosen sur- aturerange. FeCoalloyshavelargehyperfinefields(200- facedidnotdetectsecondaryphasesorrandomlydistributed 400 kOe) that would have been detected by Mo¨ssbauer nanoparticles. measurement. We can calculate the X-ray patterns ex- pected in the presence of bulk crystalline Fe impurities by superimposing the strongest peak of 0.3% elemental is caused by magnetic nanoparticles. Mo¨ssbauer mea- Fe to the measured patterns. No overlap between the surementshowsnoevidenceofironbearingnanoparticles calculated and measured Fe0.75Co0.25Sb2 X-ray patterns (eitherFeCoorFe-oxide). Suchnanoparticleswouldhave was observed (Fig. 6). Any other unknown Fe-O, Fe- a paramagnetic spectrum with different isomer shift and Co-O, Co-O, Fe-Co, Fe-Sb-Co, etc. phase with the same quadrupole splitting at room temperature, which would atomic ratio in the mixture would have been detected be detected with a 0.3% limit. Below the nanoparti- and refined by synchrotronpowder X-ray diffraction be- cle blocking temperature the field would be large, typi- causeitscontributiontothescatteringmixturewouldbe cally 500 kOe for typical oxides. Solid evidence against higher than that of Fe. Though M observed in mag- nanoparticlesorbulkextrinsicphasescomesfromenergy LL netic hysteresis loops could be caused by Fe impurities dispersivescanningelectronmicroscope(SEM)measure- of the order of the synchrotron powder X - ray diffrac- ments. Among the samples grown from several different tion detection limit, absence of hysteresis loops above 6 batchesforx=0.25,theuncertaintyinCoconcentration K strongly argues against such scenario. isx=0.04. SEMdatatakenwithresolutiondownto1.5 AnotherpossibilityisthatmagnetisminFe1−xCoxSb2 nm exclude the presence of either bulk secondaryphases 5 or embedded nanoparticles. This is because high reso- variation of the applied field. However, it is possible to lution SEM images of several randomly chosen polished ascribe the low magnetic moment in Co doped FeSb 2 crystalsandcrystallinesurfacesshownotraceofnanosize to the partial ordering of Co2+ ions. This scenario is inclusions, clusters or inhomogeneities (example shown in agreement with detailed analysis of the magnetic and in Fig. 7). The images were taken in the “composition” thermodynamic properties of Fe1−xCoxSb2.23 mode with a solid state detector consisting of paired PN Besides the obvious lattice expansion,the effect of the junctions. Thistypeofdetectorisverysensitivetoback- Co insertion is to introduce extra carriers in the system. scattered electrons which in turn are sensitive to local The carriers cause a closing of the gap by x = 0.1.23 variations in atomic number. If nano-crystallites of Fe Thus, the WFM appearance could be a consequence of or other elements were present, they would have been carrier-induced metallicity. This claim is further sup- visible as bright dots in the high magnification image. portedbydiscardinganotherwellknownscenarioforthe Taken in conjunction, our results argue against ex- WFM induction. More precisely, one can imagine that trinsic sources of WFM in Fe1−xCoxSb2. Recent muon the WFM is induced by an “inverted metal-insulator” spin relaxation measurements indicate that the WFM scenario.17,34 Inthisscenario,magneticorderexistsonly state is spread throughout the full sample volume for in the metallic phase. Furthermore, the metallicity is Fe0.7Co0.3Sb2, further supporting our results.28 a direct consequence of transition to the ferromagnetic state where a bulk moment of ∼ 1µ develops out of B small gap semiconductor with small susceptibility.17,34 V. DISCUSSION AND CONCLUSIONS However, in Fe1−xCoxSb2 for x = 0.2−0.45 a small or- deredmoment is induced. Therefore,the presence of the ExamplesofintrinsicWFMstatesinnarrowbandma- small moment excludes the “inverted metal-insulator” terialsareabundantinnature.29 Besidesnumerousoxide scenarioandleavesastheonlypossibilitythattheWFM compounds, many intermetallic systems also exhibit in- arises as a consequence of carrier-induced metallicity. trinsic weak ferromagnetism,such as YbRhSb,30 MnS,31 In conclusion, detailed structural and magnetic mea- and Yb0.8Y0.2InCu4.32 Magnetism in FeSb2 in analogy surements argue against extrinsic sources of WFM in to FeSi has been predicted by LDA+U calculations.18 Co - substituted FeSb . The ordered moment below 2 Besides the use of an external magnetic field, one in- the WFM transition for Fe Co Sb is M ∼ (0.5− 0.75 0.25 2 teresting possibility would be to induce the ferromag- 10)·10−3µ /Fe. As opposedto FeSi where the metallic B netic state by lattice expansion and band narrowing, state is caused by band narrowing of nearly ferromag- as in FeSi1−xGex.19,34 Unfortunately, isoelectronic lat- neticparentelectronicstructure,weakferromagnetismin tice expansion is limited to rather small values of x in Fe1−xCoxSb2 could be a consequence of carrier induced FeSb2−xBix. Our preliminary data show that the fer- metallic state. In order to fully understand the mag- romagnetic state is not reached for x = 0.016. As in netic structure, magnitude of moments, and mechanism Fe1−xCoxSi,ferromagneticstateisinducedwithCosub- of magnetic ordering, further neutron scattering and/or stitution in FeSb2. In both alloy systems critical tem- nuclearmagneticresonancemeasurementsareenvisaged. perature TC exhibits a characteristic peak as a function We thank Yasutomo Uemura, Paul Canfield, T. M. of Co concentration.23,24 Whereas metallicity simulta- Rice and Maxim Dzero for useful communications and neously appears with ferromagnetism in Fe1−xCoxSi at Dr. L. Rebbouh for assistance in the Mo¨ssbauer spec- x = 0.05,20 in Fe1−xCoxSb2 alloys transport and spin tral measurements. This work was carried out at the gap vanish at x=0.1 and x=0.2 respectively.23 Brookhaven National Laboratory which is operated for What could be the mechanism of the WFM in the U.S. Department of Energy by Brookhaven Science Fe1−xCoxSb2 alloys? Knowing that there is an inver- Associates (DE-Ac02-98CH10886), and at Department sionsymmetryatthe FesiteinthePnnm spacegroupof of Physics, Universit´e de Li`ege, Belgium. This work was FeSb2,wecanexcludethepresenceofthe Dzayloshinskii also supported in part by the National Science Founda- - Moriya (DM) type of interactions. This is in contrast tion DMR-0547938 (V. F. M.). to the doped FeSi where the DM interaction is believed ∗ Present address: Institut fu¨r Festko¨rperforschung, to be responsible for the WFM.33,34 A canted antiferro- Forschungzentrum Ju¨lich GmbH, D-52425 Ju¨lich, Ger- magnetism can be excluded based on the observed field many dependence of the transition temperature. That is, the ferromagnetic tail at low temperature is insensitive to 1 F. Hulliger, Nature(London), 198, 1081 (1963). and S. Arajs, Phys.Rev. 160, 476 (1967). 2 C. Petrovic, J. W. Kim, S. L. Bud’ko, A. I. Goldman, P. 4 Y. Takahashi and T. Moriya, J. Phys. Soc. Jpn. 46, 1451 C. Canfield, W. Choe and G. J. Miller, Phys. Rev. B 67, (1979). 155205 (2003). 5 G.Shirane,J.E.Fischer, Y.Endoh andK.Tajima, Phys. 3 V.Jaccarino,G.K.Wertheim,J.H.Werinick,L.R.Walker Rev.Lett. 59, 351 (1987). 6 6 K.Tajima, Y.Endoh,J.E.FischerandG.Shirane,Phys. 21 Hunter B., ”Rietica - A visual Rietveld program”, Rev.B 38, 6954 (1988). International Union of Crystallography Commission 7 L.F.MattheisandD.R.Hamann,Phys.Rev.B47,13114 on Powder Diffraction Newsletter No. 20, (Summer) (1993 ). http://www.rietica.org (1998). 8 V.I.Anisimov,J.ZaanenandO.K.Anderson,Phys.Rev. 22 A.C.LarsonandR.B.VonDreele,(ReportLAUR86-748, B 44, 943 (1991). Los Alamos National Laboratory, New Mexico, 1986). B. 9 G. Aeppli and Z. Fisk, Comments Cond. Mat. Phys. 16, H. Toby,J. Appl.Crystallogr. 34, 210 (2001). 155 (1992). 23 Rongwei Hu, V. F. Mitrovic, and C. Petrovic, Phys. Rev. 10 D. Mandrus, J. L. Sarrao, A. Migliori, J. D. Thompson, B 74, 195130 (2006). and Z.Fisk, Phys.Rev.B 51, 4763 (1995). 24 S.V.Grigoriev,S.V.Maleyev,V.A.Dyadkin,D.Menzel, 11 C.-H.Park,Z.-X.Shen,A.G.Loeser,D.S.Dessau,D.G. J. Schoenes, and H. Eckerlebe, Phys Rev. B 76, 092407 Mandrus, A. Migliori, J. Sarrao, and Z. Risk, Phys. Rev. (2007). B 52, 16981 (1995). 25 A. G´erard and F. Grandjean, J. Phys. Chem. Solids 36, 12 C.Petrovic,Y.Lee,T.Vogt,N.Dj.Lazarov,S.L.Bud’ko 1365 (1975). and P.C. Canfield, Phys. Rev.B 72, 045103 (2005). 26 J.StegerandE.Kostiner,J.Sol.St.Chem.5,131(1972). 13 A. Bentien, G. K. H. Madsen, S. Johnsen, and B. B. 27 J. B. Goodenough, J. Sol. State Chem. 5, 144 (1972). Iversen., Phys.Rev. B 74, 205105 (2006). 28 T. Uemura, privatecommunication 14 J.F.DiTusa,K.Friemelt, E.Bucher,G.AeppliandA.P. 29 L. E. De Long, J. G. Huber and K. S. Bedell, J. Magn. Ramirez, Phys.Rev.B 58, 10288 (1998). Magn. Mater 99, 171 (1991). 15 Z. Schlesinger, Z. Fisk, H-T. Zhang, M. B. Maple, J. F. 30 Y. Muro, Y. Haizaki, M. S. Kim, K. Umeo, H. Tou, M. DiTusa and G. Aeppli,Phys. Rev.Lett. 71, 1748 (1993). Sera, and T. Takabatake,PhysRev.B69, 020401 (2004). 16 A.Perucchi,L.Degiorgi, Rongwei Hu,C. PetrovicandV. 31 S. S. Aplesnin, L. I. Ryabinkina, G. M. Abramova, O. B. F.Mitrovic, European PhysicalJournal B54, 175 (2006). Romanova, A. M. Vorotynov,D. A. Velikanov,N. I. Kise- 17 V.I.Anisimov,S.Yu.Ezhov,I.S.Elfimov,I.V.Solovyev lev, and A. D.Balaev, Phys.Rev B 71, 125204 (2005). and T. M. Rice, Phys. Rev.Lett. 76, 1735 (1996). 32 A. Mitsuda, T. Goto, K. Yoshimura, W. Zhang, N. Sato, 18 A. V. Lukoyanov, V. V. Mazurenko, V. I. Anisimov, M. K. Kosuge and H. Wada, J. Phys.Chem. Solids, 63, 1211 Sigrist and T. M. Rice, European Physical Journal B 53, (2002). 205 (2006). 33 B. Lebech, J. Bernhard and T. Freltoft, J. Phys. Cond. 19 S. Yeo, S. Nakatsuji, A. D. Bianchi, P. Schlotmann, Z. Matter 1, 6105 (1989). Fisk, L. Balicas, P. A. Stampe and R. J. Kennedy, Phys. 34 V.I.Anisimov,R. Hlubina,M.A.Korotin,V.Mazurenko, Rev.Lett. 91, 046401 (2003). T.M.Rice,A.O.ShorikovandM.Sigrist,Phys.Rev.Lett. 20 N. Manayala, Y. Sidis, J. F. DiTusa, G. Aeppli, D. P. 89, 257203 (2002). Youngand Z. Fisk, Nature404, 581 (2000).