Structure and electrical resistivity of CeNiSb 3

The ternary antimonide CeNiSb 3 has been prepared from an Sb ﬂux or from reaction of Ce, NiSb, and Sb above 1123 K. It crystallizes in the orthorhombic space group Pbcm with Z =12 and lattice parameters a =12.6340(7) A˚, b =6.2037(3)A˚, and c =18.3698(9)A˚ at 193 K. Its structure consists of buckled square nets of Sb atoms and layers of highly distorted edge- and face-sharing NiSb 6 octahedra. Located between the 2 N ½ Sb (cid:3) and 2 N ½ NiSb 2 (cid:3) layers are the Ce atoms, in monocapped square antiprismatic coordination. There is an extensive network of Sb–Sb bonding with distances varying between 3.0 and 3.4A˚. The structure is related to that of RE CrSb 3 but with a different stacking of the metal-centered octahedra. Resistivity measurements reveal a shallow minimum near 25 K that is suggestive of Kondo lattice behavior, followed by a sharp decrease below 6K. r 2003 Elsevier Inc. All rights reserved.


Introduction
Rare-earth antimonides, particularly multinary phases containing transition metals, have elicited intense interest because of their important physical properties and unusual bonding. Colossal magnetoresistance has been identified in Eu 14 MnSb 11 [1,2], and relatively simple binary antimonides, such as CeSb 2 and LaSb 2 , have remarkably complex and highly anisotropic magnetic and magnetoresistance properties [3]. Pronounced f-p and f-d hybridization is believed to mediate magnetic exchange mechanisms in phases such as UMSb 2 (M=Fe, Co, Ni, Cu, Ru, Pd, Ag, Au) [4] and CeNiSb [5]. Itinerant electron ferromagnetism has been found in LaCrSb 3 and related phases [6].
An emerging feature in the structural chemistry of antimonides is the role of Sb-Sb bonds of variable strength in the formation of diverse anionic substructures such as discrete pairs (e.g., in Yb 5 In 2 Sb 6 ) [7], onedimensional chains and ribbons (e.g., in La 13 Ga 8 Sb 21 and Pr 12 Ga 4 Sb 23 ) [8,9], and most pertinent to the present work, square nets (e.g., in LaSn 0.7 Sb 2 ) [10]. These square nets appear frequently not only in antimonides, but also in other heavier pnictides, chalcogenides, and tetrelides. Application of the Zintl concept and other theoretical considerations suggest that a stable electron count for such square nets is six electrons per atom [11]. However, a 2 N ½Sb 1À net is prone to distortion, and the resulting changes in the electronic structure affect the electrical and magnetic properties.
Herein we report the synthesis, structure, and electrical resistivity of CeNiSb 3 . Its structure is related to that of REVSb 3 and RECrSb 3 [12,13] but with more varied Sb-Sb bonding interactions.

Synthesis
CeNiSb 3 can be synthesized by two methods, both of which are discussed. The single crystal used for X-ray

ARTICLE IN PRESS
diffraction experiments was isolated from the reaction of Ce (99.9%, Alfa-Aesar), NiSb (99.5%, Alfa-Aesar), and Sb (99.5%, Alfa-Aesar) which were loaded in a fusedsilica tube in a molar ratio of 1:2:5. The tube was sealed under vacuum and heated at 1123 K for 7 days followed by annealing at 873 K for 5 days. The tube was then cooled at 0.5 K min À1 to room temperature. The product consisted of excess Sb, black needles of NiSb 2 , and black tablets of CeNiSb 3 . Crystals of CeNiSb 3 up to 1 mm in length and uncontaminated with NiSb 2 could be isolated from the reaction of Ce, NiSb, and Sb in a 1:1:2 ratio under the same heating conditions but with the cooling rate reduced to 0.25 K min À1 .
Large, high-quality single crystals of CeNiSb 3 were also synthesized by a flux growth method. Ce ingot (99.95%, Ames Laboratory), Ni (99.995%, Alfa-Aesar), and Sb (99.9999%, Alfa-Aesar) were placed in an alumina crucible in a 1:2:20 ratio. The crucible and its contents were sealed in an evacuated fused-silica tube. The entire reaction vessel was heated to 1373 K where the temperature was maintained for 2 h and then cooled to 943 K at 5 K h À1 , at which point excess Sb flux was separated by centrifugation. Plate-like crystals with dimensions up to 2 Â 2 Â 1 mm 3 were mechanically separated from the alumina crucible for analysis. The crystals are stable in air and do not degrade noticeably.
Semi-quantitative SEM/EDX analysis was performed on crystals of CeNiSb 3 with use of a JEOL 840/Link Isis instrument. Ce, Ni, and Sb percentages were calibrated against standards and a Ce:Ni:Sb ratio of 1:1:3 was found.

Single crystal X-ray diffraction
A tabular crystal of CeNiSb 3 with dimensions of 0.112 Â 0.020 Â 0.048 mm and faces indexed as {100}, {010}, and {001} was mounted on a glass fiber with epoxy and aligned on a Bruker SMART APEX CCD X-ray diffractometer. Data were collected at 193 K with use of graphite monochromated MoKa radiation from a sealed tube equipped with a monocapillary collimator. SMART was used for preliminary determination of the cell constants and data collection control. Intensities were collected by a combination of three sets of exposures (frames). Each set had a different f angle for the crystal and each exposure covered a range of 0.3 in o: A total of 1800 frames were collected with an exposure time per frame of 30 s.
Data were processed with the Bruker SAINT (v. 6.02) software package using a narrow-frame integration algorithm. A face-indexed analytical absorption correction was initially applied using XPREP [14]. Individual shells of unmerged data were corrected analytically and exported in the same format. These files were subsequently treated with a semi-empirical absorption correction by SADABS [15]. The program suite SHELXTL (v. 6.12) was used for space group determination (XPREP), direct methods structure solution (XS), and least-squares refinement (XL) [14]. The final refinements included anisotropic displacement parameters for all atoms and a secondary extinction parameter. Crystallographic details are listed in Table 1. Atomic positions and displacement parameters are listed in Table 2, and interatomic distances are listed in Table 3. Further details of the crystal structure investigation may be obtained from Fachinformationszentrum Karlsruhe, D-76344 Eggenstein-Leopoldshafen, Germany (Fax: (+49)7247-808-666; e-mail: crysdata@fiz-karlsruhe.de) on quoting the depository number CSD-39122.

Electrical resistivity
The electrical resistivity of a single crystal of CeNiSb 3 along the crystallographic b-axis was measured by standard four-probe ac methods between 300 and 0.6 K using a Quantum Design PPMS instrument. Measurement on another crystal of similar dimensions confirmed that the results were reproducible.

ARTICLE IN PRESS
Extensive Sb-Sb interactions also pervade the 2 N ½NiSb 2 layer, with distances ranging from 3.0395(6) to 3.4358(6) Å (Fig. 2b). This Sb substructure can be roughly described as a three-layer stacking of 4 4 nets.
The two peripheral nets (made up of Sb2 and Sb4 atoms) are half as dense as and rotated by 45 to the intervening net made up of Sb5 and Sb6 atoms.

Structural relationships
In previously known Ce-Ni-Sb phases, the Ni atoms are in trigonal planar (CN3 in CeNiSb (ZrBeSi-type)), tetrahedral (CN4 in CeNiSb 2 (ZrCuSi 2 -type) and CeNi 2Àx Sb 2 (CaBe 2 Ge 2 -type)), or square pyramidal (CN5 in CeNi 2Àx Sb 2 (CaBe 2 Ge 2 -type)) coordination [18][19][20]. In contrast, the Ni atoms are in octahedral coordination (CN6) in CeNiSb 3 . The structures of CeNiSb 3 and CeNiSb 2 are related in that the layers of Ni-centered octahedra 2 N ½NiSb 2 in CeNiSb 3 are replaced by layers of Ni-centered tetrahedra 2 N ½NiSb in CeNiSb 2 [19], with the 2 N ½Sb net and the arrangement of Ce atoms remaining intact. A closer relationship is found with the ternary rare-earth antimonides REVSb 3 and RECrSb 3 (for concreteness, CeCrSb 3 is shown in Fig. 3a) [6,12,13]. In CeCrSb 3 , chains of face-sharing metal-centered octahedra extend along the c-direction; these chains are connected by edge-sharing in the b-direction to form a 2 N ½CrSb 2 layer parallel to the bc-plane (Fig. 3a). In CeNiSb 3 , every third metalcentered octahedron in the chains extending along the c-direction is connected by edge-sharing instead of  face-sharing (Fig. 3b). The stacking sequence of the Sb atoms along the c-direction is AB in CeCrSb 3 and ABACBC in CeNiSb 3 , or in Jagodzinski notation, h and hcc, respectively. CeNiSb 3 crystallizes in the same space group (Pbcm) as CeCrSb 3 , but its c parameter is tripled, reflecting the more complicated stacking sequence.

Bonding
Interpretation of the bonding in CeNiSb 3 is complicated by the possibility of mixed +3/+4 valency on the Ce atoms, by the pairing of Ni atoms across the shared face of the octahedra, and by the rich variety of Sb-Sb interactions within the Sb substructure. The similarity of Ce-Sb distances in CeNiSb 3 to those in CeCrSb 3 argues for Ce 3+ , and it is reasonable to assume that the Ce atoms participate, to a first approximation, in ionic interactions with the other atoms in the structure. The Ni2-Ni2 distance of 2.7214(12) Å is slightly longer than the analogous distance of B2.560 Å found in NiSb [16,17], where metal-metal bonding has long been known to be important in stabilizing its structure [21]. Magnetic measurements might also clarify the oxidation state of the Ni atoms, but the assumption of localized moments is highly suspect given the observation of itinerant electron ferromagnetism in the related series of compounds RECrSb 3 [6]. In the absence of additional experimental data, the Sb substructure can be analyzed as a starting point.
The Sb-Sb distances in CeNiSb 3 (3.0395(6)-3.4358(6) Å ) are longer than the intralayer distance (2.908 Å ) and comparable to the interlayer distance (3.355 Å ) found in elemental Sb [22]. The four-bonded Sb atoms within the 2 N ½Sb square net can be assigned oxidation numbers of -1, if the Sb-Sb interactions are considered to be one-electron bonds and two lone pairs are localized on each atom so that an octet is attained. For the Sb atoms within the 2 N ½NiSb 2 layer, however, the assumption of integral oxidation numbers breaks down. An elegant way, proposed by Jeitschko et al., to enumerate electrons within such complex Sb substructures is to apply a bond valence calculation to derive formal charges on these Sb atoms [23]. When bond orders are calculated from the equation n ij = exp[(2.80Àd ij )/0.37] (where d ij is in Å ), the formal charges of the Sb atoms in CeNiSb 3 are found to be À1.2 on Sb1, À2.5 on Sb2, À1.1 on Sb3, À2.6 on Sb4, À1.4 on Sb5, and À1.6 on Sb6. The charges on Sb1 and Sb3 are consistent with those in the simple model of one-electron bonds in a 2 N ½Sb square net (Fig. 2a). The one-bonded Sb2 and Sb4 atoms (Fig. 2b) carry the most negative charges, whereas the Sb5 and Sb6 atoms have intermediate charges between these extremes. If these formal charges are rounded off to the nearest half-integer and when the multiplicity of atomic sites is taken into account, the total negative charge of approximately À60 on the Sb atoms within a unit cell can be compensated by assuming +3 charges on the Ce atoms and +2 charges on the Ni atoms: (Ce 3+ ) 12 (Ni 2+ ) 12 (Sb 36 ) 60À or Ce 3+ Ni 2+ (Sb 3 ) 5À . It is important to appreciate that the true charges are not likely to be as extreme as implied by these formulations. The caveat about the dangers of assuming a localized electron model has already been mentioned, and significant covalent character is expected in the Ni-Sb and perhaps even the Ce-Sb bonds. It would be interesting to perform a band structure calculation to understand the bonding in more detail.  Fig. 4 shows the temperature dependence of the electrical resistivity of a single crystal of CeNiSb 3 along the b-axis. Metallic behavior is observed with a prominent curvature in the plot, which exhibits a minimum near 25 K, followed by a steep decrease at 6 K. This behavior is typical of magnetically ordered Kondo lattices with a localized f moment weakly coupled to the conduction bands. The resistivity plot bears a striking resemblance to that of CeSn 0.7 Sb 2 , a ferromagnetic layered antimonide with a similar arrangement of Ce atoms and Ce-Ce distances as in CeNiSb 3 . It is likely that the transitions in the resistivity curve can be attributed to the Ce moments. Further experiments to measure the magnetic and transport properties would be helpful in clarifying the electronic structure of CeNiSb 3 .