Table Of Content

Anisotropy in magnetic and transport properties of Fe Co Sb 1−x x 2 Rongwei Hu,1,2 V. F. Mitrović,2 and C. Petrović1 1Condensed Matter Physics and Materials Science Department, Brookhaven National Laboratory, Upton, NY 11973 2Department of Physics, Brown University, Providence, RI 02912 (Dated: February 2, 2008) Anisotropic magnetic and electronic transport measurements were carried out on large single crystals of Fe1−xCoxSb2 (0≤x≤1). The semiconducting state of FeSb2 evolves into metallic and weakly ferromagnetic by substitution of Fe with Co for x<0.5. Further doping induces structural 8 transformationfromorthorhombicPnnmstructureofFeSb2tomonoclinicP21/cstructureofCoSb2 0 wheresemiconductinganddiamagneticgroundstateisrestoredagain. Largemagnetoresistanceand 0 anisotropy in electronic transport were observed. 2 PACSnumbers: n a J I. INTRODUCTION particularlyforxvaluesclosetothe metal-semiconduc- 1 tor crossover. 1 Previous attempts to explain the unusual magnetic ] and electronic properties of the narrow-gap semicon- II. EXPERIMENTAL DETAILS el ductor FeSi1,2 were based on either the itinerant mag- - netism model, the Kondo description, the mixed valence Single crystals were grown with the high temperature r t model,orthetwo-bandHubbardmodel,andstartedwith flux method.11,12 Their structure and composition were s . Jaccarino’s interpretation of an extreme narrow-band- determinedbyanalyzingpowderX-raydiffraction(XRD) at energy-gapmodel.3−6 Fromthisworktwodifferentview- spectratakenwithCuKαradiation(λ=1.5418˚A)using m points have emerged. The first, proposed by Takahashi a Rigaku Miniflex X-ray machine. The lattice parame- - andMoriya,3 treatsFeSi as a nearly ferromagneticsemi- ters were obtained by fitting the XRD spectra using the d conductorinanitinerantelectronmodel. Thesecond,as Rietica software.13 A JEOL JSM-6500 SEM microprobe n pointed out by Aeppli and Fisk,4 takes FeSi as a transi- was used for verifying the Co concentration. Crystals o tion metal version of the Kondo insulator with localized were oriented using a Laue Camera and polished into c [ Femoments. Thisviewpointhastheorigininthenarrow rectangular bars along specific crystalline axes. Electri- gap/high density of states model that Jaccarino consid- cal contacts were made with Epotek H20E silver epoxy 2 ered and then rejected. Experimental confirmation of for a standard 4-wire resistance measurement. The di- v that model was achieved by Mandrus et al7 and Park mensions of the samples were measured by a high pre- 0 et al.8 Other analogous 3d model systems are highly de- cision optical microscope with 10 µm resolution. Mag- 5 3 sired in order to study this problem, particularly those netization, resistivity, and transverse magnetoresistance 1 of non-cubic and strongly anisotropic structures. measurements were carried out in a Quantum Design . The magnetic properties of FeSb strongly resemble MPMS-5 and a PPMS-9 varying the temperature from 1 2 0 those of FeSi.9,10 Previous investigation of FeSb2 has 1.8 K to 350 K and applying a magnetic field up to 8 shown that the magnetism can be described by two ef- 90 kOe. Polycrystalline averages were calculated from 0 fects: a thermally induced Paulisusceptibility and a low M/H ≡χ=1/3[χ +χ +χ ]. Heat capacity data was a b c : spintohighspintransitionwithinthet multipletofFe collected in the PPMS-9 instrument using a relaxation v 2g i ion in octahedral crystal field. technique. X In order to show the possible Kondo insulator to r heavy-fermion metal transition, we perturbed the FeSb a 2 III. RESULTS AND DISCUSSION electronic system by electron doping. We report the anisotropy in magnetic and electronic transport proper- A. Crystal Structure ties of Fe1−xCoxSb2 as a function of Co content. The semiconducting ground state of FeSb evolves into a 2 metalliconebyx=0.1. FurtherCosubstitutioninduces Fig. 1 shows the lattice constants and unit cell vol- weak ferromagnetism for 0.2≤x<0.5. Beyond x=0.5 ume of Fe1−xCoxSb2 as a function of nominal Co con- there is a structural transformation from orthorhombic tent, as determined by powder X-ray diffraction. As Pnnm to the monoclinic P21/c structure of CoSb and can be seen from Fig. 1, the end member structures of 2 therestorationoftheanisotropicsemiconductingground Fe1−xCoxSb2 are orthorhombic Pnnm and monoclinic statefor0.5≤x<1. Theweakferromagnetismandhigh P21/c for FeSb and CoSb , respectively. Both aˆ and ˆb 2 2 temperature Curie-Weiss behavior is due to the small axesshow a smoothlinear decreasewith increasingx for portion of localized Co ions. The large positive magne- 0≤ x< 0.5. In contrast, the a¸xis expands with increas- toresistancearisesintheFe1−xCoxSb2 electronicsystem, ing x in the same doping range. At x = 0.5, the a¸xis 2 ) 3 e(¯121.6 Pnnm P21/c 0.6 m a) u121.2 H // a ol 0.5 cell v120.8 mol) 0.4 HH //// bc Unit 120.4 mu/ 0.3 ¯) 66..46 b -3 10e 0.2 Cell constant( 3566.2...8025 a 99990000....1234 ’ M/H(-000...101 3.20 c 0.5 0.6 0.7x0.8 0.9 1.0 100 T( K2)00 300 0.0 0.2 0.4 x 0.6 0.8 1.0 b) 1.0 mol/emu)34 000...123 /F.U.)B12x=0.25 T=1 .8K FFeI1G−.xC1o:xSUb2niats aceflulnccotinosntaonftscomanpdosivtoiolunm. Pehoafsetthreanssietriioens mol) 3H/M (10a12 0.4 -3M(10--a210-6 -3 0 3 6 takes place at x = 0.5. Dashed line represents the phase u/ H(kOe) m 100 200 300 bInosuentdsahroywbsemtwoeneonclionritchaonrhgolemβb′icinanPd21m/consotrculicntiucres.tructures. -3 0e 0.5 00.1 T(K) 1 0.3 H( 0.4 M/a 01.5 0.0 50 100 150 200 250 300 350 doubles in size and a monoclinic angle β 6=90◦ emerges, T(K) itnrydiocafttihneguansittrcuecltlu.ral phase transition to lower symme- c) 5 x=0.30 mol/emu)34 0000....1234 mol) 4 00..2450 3 M (10a2 In the monoclinic region, for 0.5 ≤ x < 1, the unit u/ H/1 cell volume refers to that of a pseudo-marcasite cell. m 0.45 e 3 100 200 300 iTfihedis cbeyll −→ias′r,e−→lbat′,e−→dc′t,oatnhdeβt′r)uebyunviteccteolrlioafl CeqouSabt2io(nspseoc-f -3 H(10 T(K) the following form: −→a′ = (−→a − −→c)/2,−→b′ = −→b and M/b 2 0.20 −→c′ = (−→a +−→c)/2.14 In this doping regime as well, we 1 observe a smooth evolutionof a, b and c lattice parame- 0.0 ters with increasing x. The evolution is concurrent with 2 3 4 5 6 7 T(K) a gradualincreaseofthe monoclinic angleβ′, from90.1◦ d) 2.0 u)5 0.1 tctinohogan9tcr0eaC.nn4ot◦greas(.utFicIoincgne.a,ssdc1fodu(niIltnlfiyosorensmtu,)bi)en.nsgteiTrttguohytiVsedseligisfnaoperreadrFr’sseidvlieaenpwSet,EhndMedeemenfnooctrenirsseoetnrvdaeoCtrpeaos-l mu/mol) 1.5 0001..13 3 M (10mol/ema234 000...2345 nominal x=0.25 samples showed that the uncertainty in -3 0e 1.0 H/1 Co concentration among samples grown from different 1 H( 100 200 300 batches was ∆x=0.04. M/c0.5 T(K) The anisotropy in the evolution of lattice parameters 0.0 with Co content for 0≤x<0.5 is consistent with anin- 50 100 150 200 250 300 350 creasedoccupancyofd orbitals,directedtowardsnear- T(K) xy est neighbor Fe ions along a¸xis of the crystal.10 Mono- FIG.2: M/Hv.s.T at H =1000 Oealong3crystallineaxes. Some data of the series were omitted for clarity: (a) mag- clinic distortion at x = 0.5 increases the a¸xis and de- netization of FeSb2 along all three axes. (b)Magnetization creases the overlap of d orbitals. The reduced over- xy along aˆ axis. Insets show the 1/χ and also the hysteresis lap results in an enlargement of the gap within the t multiplet15 and causes a change in electronic transpo2rgt loop of Fe0.75Co0.25Sb2. (c) Magnetization along ˆb axis dis- plays weak ferromagnetic transition at low temperatures for properties. The electronic system gradually evolves to- 0.2 ≤x≤ 0.45. (d) a¸xis magnetization resembles those of aˆ wards a semiconducting ground state of CoSb2 with the axis but shows no transitions at low temperatures. increase of Co concentration from 0.5≤x<1. 3 TABLE I:Parameters of thefits tothe polycrystalline average of M/H data for 0≤x<0.5. χ(T)=χNB+χCW x ∆χ(K) W(K) p χ0(10−4emu/mol) µeff(µB) Θ(K) µeff/Co(µB) Tc(K) Ps(10−3µB) 552 1.6 1.3 0.03 0 0 0 425 310 1.4 0.04 0.05 531 235 2.1 0.3 0.41 −23 0.10 511 471 2.5 1.6 0.79 −71 0.20 4.5 0.14 110 0.70 6.6 1.49 0.25 3.3 0.24 104 0.99 6.0 1.73 0.30 3.0 0.32 77 1.08 6.8 1.87 0.40 1.5 0.54 17 1.35 7.3 1.19 0.45 1.5 0.54 16 1.20 6.4 0.75 B. Magnetic Properties The magnetic susceptibility of FeSb2 can also be described by the narrow-band-small-gappicture: For x = 0, the temperature dependence of the susceptibility is qualitatively the same for H||aˆ,ˆb, and cˆ χ = 2Nµ2Bpexp(∆χ/T)(exp(W/T)−1) +χ NB 0 W(1+exp(∆ /T))(1+exp((∆ +W)/T)) as shown in Fig. 2(a). A diamagnetic to paramagnetic χ χ crossoveratT inthevicinityof100Koccurswhenafield is applied parallel to the câxis. This crossoveris absent Here χNB is the Pauli susceptibility of an itinerant when fields are appliedparallelto the aˆ andˆb axes. Fur- electron system with a density of states N(E) consist- thermore, for H||aˆ,ˆb below 100 K the susceptibility is ing of two narrow bands of width W separated by a gap E ≡ 2∆ , so that the chemical potential is de- temperatureindependentandcanbedescribedbyapara- g χ magnetic term χ =χ =2×10−5 emu/mol. These ob- fined as µ = W +∆χ. The density of states is given by a b N(E) = Np/W, where N denotes the number of unit servations contradict previous measurements of FeSb ,9 2 cells and p the number of states per unit cell. In this thatreportedadiamagnetictoparamagneticcrossoverin picture,χ isthetemperatureindependentparamagnetic thevicinityof100Kforfieldsappliedalongallthreeaxes. 0 susceptibility. Byfittingthismodel,wefindtheallowable Our measurements were performed on large single crys- values of parameters: 1.3<p<1.4, 1.6K <W <310K, talsampleswhicheliminatedtheneedforsubtractingthe 850K < E < 1100K. The result agrees with Ref.9 but background,whichisoftendifficulttoestimatewithhigh g differs inapositive Pauliterm,χ =3×10−6−4×10−6 certainty. Subtraction of the background was necessary 0 emu/mol, instead of a diamagnetic term. in Ref. 9 due to very small mass of the FeSb samples. 2 At high temperatures for 0 < x ≤ 0.45, the magnetic Consequently, the intrinsic susceptibility of FeSb deter- 2 susceptibility along all three crystalline axes can be de- mined on large single crystals, differs from Refs. 9 and scribedbytwoterms: athermallyinducedparamagnetic 10 and is shown in Fig.2(a). moment with progressively smaller gap ∆ , and a grow- In what follows, we explore to what extent different χ ing Curie-Weiss term, due to Co substitution: models can account for the observed paramagnetic mo- mentandthetemperaturedependenceofthesusceptibil- ity in Fe1−xCoxSb2. In the free-ion model the suscepti- χ(T)=χ +χ bility χ is given by Jaccarino et al2 as, CW NB J(J +1) 2J +1 where χ =Ng2µ2 +χ . FI B 3kBT 2J +1+exp(∆χ/kBT) 0 Nµ2 eff χ = This model describes thermal excitations from low to CW 3k (T −Θ) B high spin states, separated by a spin gap ∆ . More- χ over in the simplest case, two spin states can be viewed Results of fits for 0 ≤ x < 0.5 are shown in Table I. astheJ =S =0spinsingletandtheS =1tripletstate. Thegap∆χdecreaseswithincreasingdopingasxreaches Fitting the expression to the polycrystalline average of 0.2. For 0.20≤x≤0.45 the temperature dependence of the magnetic susceptibility of FeSb and setting J = S, the magnetic susceptibility above 180 K is characterized 2 we obtain χ = 4.5×10−6emu/mol, ∆ = 527 K and solelybytheCurie-Weisstermandtemperatureindepen- 0 χ S = 0.24. The rather small value of S implies that the dent Pauli term χ0. free-ion model may not be the explanation for the ther- The low temperature magnetic properties of mally induced paramagnetism at high temperatures, or Fe1−xCoxSb2 are anisotropic. For H||aˆ and ˆb, we that a small portion of Fe spins in the S=0 low spin d4 observe a large increase of M/H values, a signature ground state(t ,S=0) takes part in the spin-state tran- of ferromagnetic transition. Consequently, hysteresis 2g sition. loops, shown in the inset to Fig. 2 (b), are observed for 4 600 1000 100 Pnnm P21/c 500 10 400 0 5 /aa3 1 x T(K)300 0 0.05 200 0.10 0.20 0.40 100 0.50 0.75 1 0 1000 0.0 0.2 0.4 0.6 0.8 1.0 100 x 10 FIG. 4: Spin and transport gaps as a function of Co concentration. The Curie-Weiss behaivor is widely identified in the 0 35 1 series. Both spin andtransport gapsareshrinkingduetoCo b /b impurity states, which likely reside in thegap. is diamagnetic in the whole experimental temperature range. As Fe is in the nonmagnetic 3d4 configuration, 0.1 we assume that localized moment arises solely from the 10000 substituted Co atoms, and hence Curie-Weiss term. To evaluate the high temperature moment per Co 1000 atom, we fit the polycrystalline average of the magnetic susceptibility for temperatures ranging from 180 to 350 100 K with the Curie-Weiss law. We find the moment of 0 c35 µ/Co to be 1±0.35µB for all the ferromagnetic x (Tab. /c 10 I). This suggests the existence of polarized moments of Co2+ with low spin and quenched orbital momentum 1 in the lattice. Therefore, we conclude that the weak ferromagnetism is due to the ordering of 0.2 % of Co2+ moments. The fact that we were unable to detect any 0.1 hysteresisloops aboveT and the absence of any sample c 10 100 dependence among sample within one batch of crystals, T(K) as well as among crystals from different batches, argues against extrinsic sources of weak ferromagnetism in FIG. 3: Normalized resistivity along all three axes as a function oftemperatureanddifferentdopinglevels, xdenotedby Fe1−xCoxSb2. different shapes and colors. Fe1−xCoxSb2 becomes metallic alongallaxesforx>0.05. Withincreasingx,themetallicity is lost in a crossover, where resistivity is nearly independent C. Electronic Transport of temperature. Further increase of T leads to metallic behavior again along aˆ andˆb. However, resistivity along câxis evolves gradually from metallic to semiconducting. Below 40 Whereas the electronic transport of FeSb is strongly 2 K for x=0and 0.05 samples theimpuritytails up to5 Ωcm anisotropic,thereismuchlessdependenceon”direction” are observed. for Fe1−xCoxSb2. Figure 3 shows the normalized resistivity as a function of temperature along all three axes. Absolute values of the resistivity at T =350 K for sam- 0.2≤ x ≤ 0.45 (H||ˆb), as well as for x = 0.25 (H||aˆ) at ples with doping up to x = 0.4 are listed in Tab. II. T=1.8K. The saturation moments Ps are listed in Table For x = 0 (FeSb2), there is metallic behavior above 40 I. The Curie temperatures, determined by extrapolating K along the ˆb axis, and activated behavior below that thesteepestslopeoftheM/HcurvetoM =0,varyfrom temperature. On the other hand, FeSb is semiconduct- 2 5.9 to 7.2 K (Tab I.). This correspondsto the anomalies ing for current applied along aˆ and a¸xes. With a small in the heat capacity measurements (Fig. 6). CoSb amountofdoping (x=0.1),the resistivityexhibits good 2 5 metallicbehavioroverthewholetemperaturerange,and ˆb axis transport we observe up to two orders of magni- showing very small defect scattering (ρ ∼ 10µΩcm) at tudeofpositivemagnetoresistanceattemperaturesinthe 0 1.8 K, the lowest measured temperature. vicinity of the metal to semiconductor crossovertemper- For 0.1 < x < 0.4 the electronic transport is metallic ature. Away from the crossover temperature the effect for current applied along all three principal crystalline diminishes, below 5 K and in the vicinity of room tem- axes. Metallicity induced by Co substitution is sup- perature the applicationofa magnetic field has no effect pressed as the system evolves towards the region of the on the ˆb axis transport. structural phase transition for x ≥ 0.5, due to mono- For x = 0.1, this strong effect is more isotropic and clinic distortion that causes doubling of the a¸xis and the is observedfor current applied along all three crystalline consequent reduction of the overlap of dxy orbitals. For axes. However for I||ˆb, the effect is strongest. Below 0.5 ≤ x ≤ 1(P21/c structure) a small anisotropy ap- 10 K, the resistivity increases up to 2.5 orders of magni- pears again, but it has different character comparing it tude in an applied field of H = 90 kOe. The transverse to0≤x<0.5(Pnnmstructure). Thestructuralchange magnetoresistance (MR), defined as is likely one origin of that difference. More theoretical and experimental band structure work would be useful [ρ(90kOe)−ρ(0)] MR= , to shed light on the anisotropy of electronic transportof ρ(0) Fe1−xCoxSb2(0.5≤x≤1). For the semiconducting samples, we identify a regime reaches36000%atT =1.8K and∼ 30 %atT =300K of thermally activated resistivity above 40 K described in 90kOe. This effect is comparable to the colossalmag- by netoresistanceobservedbelowroomtemperatureinman- ganites. Furthermore,itis muchhigheratroomtemper- ∆ρ ρ=ρ0e2kBT. ature than in manganites. The origin of the magnetore- sistanceinFeSb (x=0)ismostlikelyduetothelargemo- 2 Here ∆ is the transport energy gap. The activation ρ bility (µ) of carriers. The mobile carriers in a magnetic energyforthe transportgapis obtainedattemperatures fieldmoveincyclotronorbitsspecifiedbyω τ ∼µH >1. c above40K.Thespingapsareobtainedfromthenarrow- Therefore, a strong positive magnetoresistance effect is band-small-gap model of the polycrystalline magnetic expected in even a moderate field.9 susceptibility. For x=0.05 the average value of ∆ from ρ Magneticisothermsdeviatefromthequadraticrelation ρa(T) and ρc(T) is shown(they agree within measure- for the ordinary metallic case, i.e. ∆ρ(H)/ρ ∝ H2 and 0 ment error∼10%), since ρ is metallic. For 0.5 ≤ x < 1 b there is a poor Kohler scaling for Fe Co Sb . The 0.9 0.1 2 the value of ∆ from ρ is shown since ρ and ρ are ρ c a b spin fluctuation mechanism proposed by Y. Takahashi metallic. Near structural phase transition, the value of and T. Moriya 3 can be discarded based on the sign of ∆ from ρ is shown since ρ and ρ are metallic. A ρ a b c the effect. Mainly because the spin fluctuation theory discrepancy between the spin and the transport gap val- predictsnegativemagnetoresistanceduetothethermally ues is evident. A possible difference between the gaps in charge and spin excitation channels has also been observed in pure FeSb 9 as well as in FeSi16,17. 2 TABLE II: Values of resistivity Fe1−xCoxSb2(0≤x≤0.4) at 350 K. Fe1-xCoxSb2 x=0 a x ρa(mΩcm) ρb(mΩcm) ρc(mΩcm) 1000 b 0 0.79 0.27 0.63 c 0.05 0.74 0.23 0.26 100 x=0.1 a 00..12 00..3996 00..2509 00..2648 cm) 10 b 0.3 1.38 0.50 0.69 m c ( 0.4 1.80 1.04 1.56 1 In FeSb application of a magnetic field induces a 0.1 2 large anisotropic positive magnetoresistance for current applied along the highly conductive ˆb axis.9 This effect 0.01 is strongly enhanced in Fe1−xCoxSb2, particularly for 10 100 highlymetallicsamples. As anexample,inFig. 6theef- T(K) fectinFe Co Sb isshown. Inaddition,magnetoresis- 0.9 0.1 2 tance for pure FeSb is plotted for comparison. Whereas FIG. 5: Large magnetoresistance for x = 0 and moderately 2 magneticfieldhaslittleeffectonthesemiconductingelec- doped(x=0.1)highlymetallicsamples. Solidandopensym- tronic transportalong the aˆ and câxes in FeSb , for the bols denote datain 0 and 90 kOeapplied fields, respectively. 2 6 suppressed spin fluctuation amplitude. In addition, the TABLEIII:Sommerfeld coefficient andeffecitvemassof car- observedtemperature dependence ofmagnetizationdoes riers. not contain T3/2 and T5/2 terms that arise from spin 2 ∗ wave and spin wave - spin wave interactions. x γ(mJ/molK ) m (me) 0.20 25.9 1.35 0.25 27.6 1.33 0.30 29.96 1.33 D. Heat Capacity 0.40 22.9 0.95 0.45 22.4 0.89 The heat capacity C /T as a function of T2 below 10 p K is shown in Fig. 6. We observe a weak but clear sig- natureofaferromagnetictransitioninthevicinityofthe hancement, we make the free-electronassumption. That T ,asderivedfrommagnetizationmeasurements. Byin- C is, tegratingC /T overT underthis anomaly,the magnetic p entropy contribution per Co is evaluated. For x = 0.3, we find ∆S =0.35%kBln2, in good agreement with the γ = π2kB2N(EF) = kB2kFm∗ results obtained from the magnetization measurement. 3 3h¯2 The end members of the Fe1−xCoxSb2 series have negligible electronic specific heat γ = 9.2× 10−3 mJ/mol with k = (3π2n)1/3. The carrier density is denoted by F K2. This implies an negligible density of states at the n and is estimated to be equal to the Co concentration Fermi level, i.e. an insulating state. The result agrees assuming one itinerant electron per Co2+. This model well with optical spectroscopy measurements that show impliesonlyasmallenhancementofcarriermassanddis- a negligible Drude weight of σ(ω) at low frequencies.18 agrees with the Kondo Insulator model. However, Hall The changes in γ as a function of x are correlated constantmeasurementswouldbeusefultodeterminethe with the evolution of the magnetic properties, since the precisecarrierdensityandsoprovideadefinitivetestfor saturation moment dependence for samples in the ferro- the applicability ofthe KondoInsulatormodel. Further- magnetic region reaches a maximum value for x = 0.3. more, these measurements will help shed more light on Within the Kondo insulator framework, a metallic state the origin of the colossal magnetoresistance in this sys- with a large carrier mass enhancement is generated by tem. a carrier induced closing of the hybridization gap. Con- sequently, the carrier mass enhancement leads to large γ valueduetoarenormalizedbandstructure. Theγ values observedin Fe1−xCoxSb2 arecomparableto those inthe IV. CONCLUSION ferromagnetic regime of FeSi1−xGex16 and FeSi1−xAlx, Fe1−xCoxSiaswell.17,19 Toestimatethecarriermassen- In summary, the Fe1−xCoxSb2 electronic system ex- hibitstwoconsecutivemetaltosemiconductorcrossovers. The first is caused by electron doping of the semiconducting ground state of FeSb for x = 0.1. The second 2 is induced by a structural change for doping levels in 50 the vicinity of x = 0.5. The density of states at the Fermi level increases due to the increased Co concentration, as shown in the rise of the electronic specific heat. x=0.30 The weak ferromagnetism induced by the polarization 45 of a small portion of Co2+ ions arises in the metallic 2 K) 0.25 state for 0.2 ≤ x < 0.5. A colossal magnetoresistance ol m as largeas 36000%is observedin the vicinity of the first J/ 0.20 m metaltosemiconductorcrossover. Theoverallproperties p/T(40 0.40 of Fe1−xCoxSb2 can be understood in the narrow-band- C small-energy-gapframeworkwhereCo substitutiongives 0.45 riseto impuritystates inthe gap,metallicity,andCurie- 35 Weiss behavior. This is further supported by the recent LDA+U calculation of Lukoyanov on FeSb , which in- 20 2 0% 100% dicates sharp peaks in the DOS at the edge of a small 0 20 40 60 80 100 gap.20 Possible consequence of the presence of magnetic T2(K2) Fe or an unknown ferromagnetic impurity would imply that the value of the direct Coulomb repulsion param- FIG. 6: Cp/T as a function of T2. Cusps correspond to fer- eter U in ref. 20 does not exceed the critical value of romagnetic transition. Co doping gives rise to substantial U =2.6eV for all values of x in Fe1-xCoxSb2 and that C increase in gamma. the weak ferromagnetism is extrinsic.20 7 V. ACKNOWLEDGMENTS This work was carried out at the Brookhaven National Laboratory, which is operated for the U.S. Department WethankDonavanHall,T.M.Rice,P.C.Canfield,S. of Energy by Brookhaven Science Associates (DE-Ac02- L. Bud’ko, Myron Strongin and M. O. Dzero for useful 98CH10886). This work was supported by the Office of communicationandJohnWarrenforSEMmeasurement. BasicEnergySciencesoftheU.S.DepartmentofEnergy. [1] G. Föex, J. Phys. Radium 9, 37 (1938). Earths, Vol. 12, Elsevier, Amsterdam, (1989). [2] V. Jaccarino, G. K. Wertheim, J. H. Wernick, L. R. [13] Hunter B., ”Rietica - A visual Rietveld program”, Walker,and S.Arajs, Phys. Rev. 160, 476 (1967). International Union of Crystallography Commission [3] Y.TakahashiandT. Moriya, J.Phys.Soc.Jpn. 46,1451 on Powder Diffraction Newsletter No. 20, (Summer) (1979). http://www.rietica.org (1998) [4] G.AeppliandZ.Fisk,CommentsCondens.MatterPhys. [14] EinarBjerkelundandArneKjekshus,ActaChem.Scand. 16, 155 (1992). 24,3317-3325(1970). [5] C. M. Varma, Phys. Rev.B 50, 9952 (1994). [15] J. B. Goodenough, J. Solid State Chem. 5, 144 (1972). [6] V.I.Anisimov,S.Y.Ezhov,I.S.Elfimov,I.V.Solovyev [16] S. Yeo, S. Nakatsuji, A. D. Bianchi, P. Schlottmann, Z. and T. M. Rice, Phys.Rev.Lett. 76, 1735 (1996). Fisk,L.Balicas,P.A.Stampe,andR.J.Kennedy,Phys. [7] D.Mandrus, J.L.Sarrao, A.Migliori, J.D.Thompson and Rev. Lett.91, 046401 (2003). Z.Fisk, Phys. Rev.B 51, 4763 (1995). [17] J.F.DiTusa, K.Friemelt, E.Bucher, G.Aeppli, [8] C.-H.Park,Z.-X.Shen,A.G.LoeserandD.S.Dessau,Phys. A.P.Ramirez, Phys. Rev.B 58, 10288 (1998). Rev.B 52, 16981 (1995) [18] A. Perucchi, L. Degiorgi, R. Hu, C. Petrovic, and V. [9] C.Petrovic,J.W.Kim,S.L.Bud’ko,A.I.Goldmanand Mitrović, Eur. Phys. J. B 54, 175 (2006). P.C. Canfield, Phys. Rev.B 67, 155205 (2003). [19] M. A. Chernikov, L. Degiorgi, E. Felder, S. Paschen, A. [10] C.Petrovic,Y.Lee,T.Vogt,N.Dj.Lazarov,S.L.Bud’ko D. Bianchi, H. R. Ott, J. L. Sarrao, Z. Fisk, and D. and P. C. Canfield, Phys. Rev.B 72, 045103 (2005). Mandrus, Phys. Rev.B 56, 1366 (1997). [11] P.C. Canfield, Z. Fisk Phil. Magaz. B 65, 1117 (1992). [20] A. V. Lukoyanov, V. V. Mazurenko, V. I. Anisimov, [12] Z. Fisk, J. P. Remeika, in: K. A. Gschneider, J. Eyring M.Sigrist,andT.M.Rice,Eur.Phys.J.B53,205(2006). (Eds.),HandbookonthePhysicsandChemistryofRare