We report the observation of an unexpected magnetic transition in the Kondo semimetal CeNiSn under uniaxial pressures. For Pʈa, both the specific heat C and magnetic susceptibility decrease steadily with increasing P. By contrast, once Pʈb and Pʈc exceed 0.13 GPa, C(T) exhibits a distinct jump at 4 K followed by a peak at 3 K. The concomitant peaking in (T) for PʈBʈc is indicative of an antiferromagnetic transition. Pressure can turn a magnetically ordered state to a nonordering state in

more »

... y correlated electron systems such as heavy-fermion compounds and transition-metal oxides by tuning the electronic couplings and band width. 1 For example, in a cerium-based antiferromagnet at ambient pressure, the intersite Ruderman-Kittel-Kasuya-Yoshida ͑RKKY͒ interaction dominates the on-site Kondo interaction. 2 Under pressure, however, the Kondo interaction is enhanced more strongly than the RKKY interaction, leading to a magnetic instability. The approach from a nonmagnetic state to the magnetic instability under pressure has been realized only in certain ytterbium-based compounds. 3 Recently, much interest has focused on anomalous phenomena at the critical pressure, i.e., non-Fermi-liquid behavior 4 and unconventional superconductivity. 5 Most Ce-based compounds being away from the critical point possess a metallic ground state. However, CeNiSn forms a pseudogap in the density of states at the Fermi level. 6 The magnitude of the energy gap was derived from nuclear magnetic resonance and specific-heat measurements to be 14 and 21 K, respectively. 7,8 The anisotropic behavior in magnetic and transport properties originates from the strong c-f hybridization of the Ce 4 f states with the Ni 3d and Sn 5p band states in the orthorhombic structure. 9 The magnetic susceptibility along the a axis, a , is much larger than b and c , and exhibits a pronounced peak at 12 K. 6 Below this temperature, dynamic antiferromagnetic correlations develop, 10 but no transition into static magnetic order occurs even at 10 mK. 11 A systematic study has shown that the gap formation in CeNiSn is very sensitive to the degree of c-f hybridization. 9,12-15 Partial substitution of Cu, Pd, and Pt for Ni strongly suppresses the pseudogap, 9 leading to a longrange magnetic order at low temperatures for 8 at. % Cu and 33 at. % Pt. 12,13 These substitutions expand the unit cell volume, which may weaken the c-f hybridization. Under a hydrostatic pressure, by contrast, the c-f hybridization should be strengthened. However, the pseudogap is also suppressed as was indicated from the decrease in the absolute value of the Hall coefficient. 14 This pressure effect suggests that the carrier concentration increases together with the recovery of the density of states at the Fermi level. Furthermore, suppres-sion of antiferromagnetic correlations upon applying pressure was deduced from a neutron inelastic-scattering experiment. 15 It should be recalled that noncubic heavyfermion compounds such as CeCu 5.8 Au 0.2 ͑Ref. 16͒ and UPt 3 ͑Ref. 17͒ exhibit significantly different responses to an uniaxial pressure from that to a hydrostatic pressure. Therefore, one expects an anisotropic response of the pseudogap in CeNiSn to a uniaxial pressure. In this paper, we report the observation of a magnetic ordering in CeNiSn under uniaxial pressures by the measurements of the specific heat and magnetic susceptibility. The single crystal of CeNiSn was prepared by a Czochralski method using a hot tungsten crucible in a radio-frequency induction furnace. Details of the preparation methods and characterization of the crystals were described elsewhere. 18 The specific heat under uniaxial pressures up to 0.4 GPa was measured using an ac method in the temperature range 1.7 рTр20 K. A disk-shaped 3.0 mmϫ0.2 mm sample was cut perpendicular to the principal axis. The sample was sandwiched between two Cu plates, on which a thermometer and a heater were mounted, respectively. A pair of pistons was made of ZrO 2 with a rather low thermal conductivity. To achieve better thermal isolation, diamond powder was placed between the Cu plate and the piston. The pressure was determined by the known pressure dependence of the superconducting transition temperature T c ( P) of indium. The transition was measured by the mutual induction method with coils outside the pressure cell. The contribution of the Cu plates, thermometer, and heater to the total heat capacity was determined in separate measurements without the sample. We found that this contribution is 10-20 % of the total heat capacity for 2рTр10 K, and is essentially independent of pressure up to 0.3 GPa. The absolute value of the specific heat of the sample was yielded by comparing with the value obtained by an adiabatic method at ambient pressure. 10 The magnetic susceptibility under uniaxial pressure for B ʈ P was measured by a superconducting quantum interference device magnetometer in the range 2рTр10 K. The pressure cell and pistons were, respectively, made of hardened Cu-Be alloy and ZrO 2 . The pressure was determined by the known dependence of T c ( P) for tin. Details of the experimental setup are described in Ref. 19.