Magnetic structure of the heavy-fermion compound U2Znt7

The phase transition of U2zni7 at 9. 7 K has been investigated by neutron powder diffraction. The transition corresponds to the onset of antiferromagnetic order where the U moments are oriented antiparallel to their nearest neighbors within the basal planes and the near neighbor along the c axis of this rhombohedral compound. At 5 K, the ordered moments lie within the basal planes and are of magnitude (0. 8+0. 1) p, &, which is substantially below the paramagnetic moment of 2. 25', &/U atom given by high-temperature susceptibility data.

8093 Zurich, Switzerland (Received 26 September 1985) The phase transition of U2zni7 at 9.7 K has been investigated by neutron powder diffraction. The transition corresponds to the onset of antiferromagnetic order where the U moments are oriented antiparallel to their nearest neighbors within the basal planes and the near neighbor along the c axis of this rhombohedral compound. At 5 K, the ordered moments lie within the basal planes and are of magnitude (0.8+0.1)p, &, which is substantially below the paramagnetic moment of 2.25', &/U atom given by high-temperature susceptibility data.
The list of "heavy-fermion" systems, which display extraordinarily large ( & 400m, ) effective conduction electron masses as deduced from the low-temperature specific heat, now includes both superconductors and metals remaining paramagnetic to the lowest temperatures.
Recent magnetic susceptibility and specific-heat measurements suggest that there are also heavy-fermion systems which undergo transitions to magnetically ordered ground states. ' In this paper, we present the first neutron-scattering determination of magnetic order in a heavy-fermion system, U2Zni7. The essential results are as follows. First, the ordering, which sets in at T~= 10 K, is exceedingly simple, with the magnetic unit cell identical to the nuclear unit cell, and the moments associated with the two U atoms in the primitive unit cell oriented in opposite directions. Second, the ordered moment is (0.8+0.1)p,s/U atom which is well below the moment of 2.25@, s /U atom deduced from the high- For the structure determination at 15 K ( ) T~), a pyrolytic-graphite analyzer in the (004) setting was used in order to optimize the resolution at higher scattering angles. The collirnations were 20'-open-40'-20' for in-pile, monochromator-sample, sample-analyzer, and analyzer-detector, respectively.
The resulting diffraction pattern showed a number of weak impurity peaks in addition to those characteristic of U2Zni7. The major impurity was identified as Zn, estimated to be about 5'/0 by weight.
The remaining peaks were an order of magnitude weaker (with intensities «1% of the strong U2Zn~7 peaks) and could not be identified. The d spacings are listed in the caption to Fig. 1, and do not correspond to those of UZni2 or e-U.
At room temperature, U2Zn&7 (Ref. 3) has the Th2Zn~7type structure, which is one of a series of ordered structures [e.g., Pu3Zn22, UZnt2 (Ref. 5)] which can be derived from that of CaCu5. In U2Zni7, Zn occupies the Cu sites and U is ordered on two-thirds of the Ca sites, the remaining onethird being replaced by pairs of Zn atoms about 2.6 A apart. ' The ordered structure has rhombohedral symmetry, space group R3m, with hexagonal lattice constants v 3a and 3c with respect to those of the parent CaCu5-type cell. Because the structural parameters for U2Zni7 near its 10-K transition were unknown, we carried out a Rietveld analysis of the data sho~n in Fig. 1. Regions around the impurity peaks and Al reflections from the sample holder were excluded from the refinement, and background contributions were estimated by interpolation between values obtained by averaging over regions where no Bragg peaks were present. Table I displays the lattice parameters and atomic coordinates given by the profile refinement. In the upper frame of Fig. 1, the solid line represents the calculated profile which best fits the data, while in the lo~er frame, the difference between the calculated and observed profiles is shown.
The results are in exce11ent agreement with a room-temperature x-ray scattering determination of the structure. 3 Belo~10 K some very weak additional scattering at some of the low-angle nuclear peak positions was observed. To gain more intensity and thus allow adequate counting statistics to be obtained in a reasonable period of time, the  analyzer setting was changed to (002) and the collimation relaxed to 40'-80'-40'-40'. Figure 2 shows the low-angle diffraction data for temperatures well above (15 K) and well below (5 K) the peak observed at 10 K in the specific heat for U2Zni7. ' The allowed nuclear reflections are labeled by hexagonal indices, with arrows indicating their positions.
The important features of the data are (1) that no superlattice peaks appear at low temperature, and (2) that the intensity at some of the nuclear Bragg peak positions increases upon lowering T through T~. We note that the (012) nuclear reflection is allowed by symmetry, but is accidentally weak. Polarized neutron-scattering measurements on a single crystal of U2Zni7 also demonstrate the magnetic origin of the enhanced Bragg scattering at low temperatures. The inset to Fig. 1 shows the temperature dependence of the magnetic contribution to the peak heights of the (101) and (012) reflections, from which the transition temperature is seen to be about 10 K.
Because the magnetic scattering occurs only at nuclear peak positions, the magnetic unit cell is the same size as the chemical cell and contains two U atoms in the primitive unit cell at + 00z, where z is very close to T (see Table I). Since ferromagnetism is precluded by the susceptibility data, ' the two moments must be antiferromagnetically coupled, which leaves only their direction and magnitude to be determined. The latter requires knowledge of the scale factor relating Bragg intensities measured in counts per minute to scattering cross sections expressed in barns; this factor was determined from a refinement of the high-intensity nuclear reflections obtained for 43 & 28 & 66' performed by using the known scattering lengths of U and Zn, and the structural parameters obtained from the Rietveld analysis of the higher-resolution, 15-K data described above. The subsequent intensity calculations (see Table II) show that the moments have a magnitude of 0.8+O.lp, s/U atom and lie in the basal planes. From power diffraction data alone, it is not possible to specify the direction in the planes. The magnetic structure is illustrated in Fig. 3. In addition to antiferromagnetic coupling between near-neighbor c-axis pairs of moments 4.36 A apart, a given moment is coupled antiferromagnetically to three near neighbors in the basal planes at a distance of 5.16 A.
While there are many other U compounds which order magnetically, " U2Zni7 is unusual because of its low U den-  (101)  that large magnetic fluctuations persist at T 0. In particular, even belo~T~, there is a large "electronic" contribution y T to the specific heat, as well as a large Pauli-type sus-ceptibility X0=0.9 emu/mole U atom. It is amusing to note that i(pa, rrpo)/p, , frl'=0. 4, which is the mean square fluctuating moment pf normalized to p, , 'ff, has the same value as y(T 0)/y(T = Tw). Thus, U2zni7 behaves as an itinerant antiferromagnet where a gap develops only over a fraction 1y(T 0)/y(T= Tg) of the Fermi surface. " It may also be appropriate to compare U2ZnI7 to singlet ground-state magnets, where the ordered moment p,o is also suppressed relative to p, ,ff. '" Such an analogy is particularly attractive because U2ZnI~is suspected to be a dense Kondo system, where the interactions between the conduction electrons and moments localized near the U atoms yield singlet TABLE II. Observed and calculated nuclear and magnetic integrated intensities from U2zn&7 at 15 and 5 K. Nuclear intensities are based upon the parameters listed in Table I, and magnetic intensities are calculated for the antiferromagnetic structure illustrated in Fig. 3  As far as the high-temperature (T ) T~) specific heat' is concerned, conventional Schottky anomalies need not be found because 4 is generated dynamically, via interactions with the conduction electrons, rather than being due to fixed crystal-field levels. In conclusion, we caution that awhile it seems useful to consider U2Zn~7 as a collection of single Kondo impurities, there is no doubt that a solution of the lattice problem'~is required to understand several important properties, notably the sensitivity to impurities and temperature dependence of the resistivity.