NEUTRON SCATTERING EXPERIMENTS ON THE MAGNETISM IN Cu-Mn SINGLE CRYSTALS

We have carried out neutron polarization analysis experiments, and unpolarized neutron scattering experiments, on annealed single crystal Cu-Mn samples of 5, 10, 15 and 25 atomic percent Mn. Peaks in the magnetic diffuse scattering are found at the [1, h ± 5. 0]-type positions in reciprocal space, where 6= 0.34, 0.29, 0.25 and 0.19 for the 5, 10, 15 and 25 percent Mn samples respectivly. This magnetic scattering is symmetry-related to the diffuse nuclear scattering centered at the LI, h, 0]-type positions arising from atomic short range order. The intensity of this diffuse magnetic scattering in all of the alloys decreases smoothly with temperature, approaching zero in the range 300-350 K. The spin-spin correlation range, as judged from the width of the magnetic diffuse peaks, also decreases smoothly with temperature. Neither the width nor the intensity of this magnetic scattering gives any hint of a freezing transition at lower temperatures.

. In spite of the fact that a large number of neutron scattering experiments, (4,5,6,7,8) of increasing sophistication, have been carried out on these alloys, a full understanding of the static and dynamic magnetic structure of these materials has not yet evolved. Fundamentally, we are faced with the problem of understanding the origin of the cusp in the low-field magnetic susceptibility (9) occuring at some temperature dependent upon Mn concentration, in the absence of the development of long-range antiferromagnetic order at low temperatures. Since there is no abrupt change in the entropy at the temperature at which the susceptibility cusp occurs, now commonly called the freezing temperature T f , the low temperature magnetic structure of these alloys is 2.
thought to resemble the static (metastable) structure of a glass. This spin glass or mictomagnetic behavior now appears to be a general characteristic of many more or less disordered alloys [such as Au(Fe), Ag(Mn), (La, Gd )AJU]> irrespective of the concentration of magnetic atoms.
It is clear that the macroscopic magnetic properties of these alloys depends intimately upon the spatial distribution of atoms in the crystals; that is, on the short range atomic order. This fact is particularly apparent from the drastic changes in the magnetic susceptibility upon heat treatment. For example, in Cu^cMnpc the height of the cusp in the low field susceptibility is a factor of 4 larger in an annealed sample than in a quenched sample (9). It is known from neutron scattering experiments that the short range atomic order increases with annealing (10). This suggests that short range atomic order, or inhomogeneity, may be a prerequisite to spin glass behavior. In order to determine, more precisely, the positions in reciprocal space where the magnetic diffuse scattering peaks, we have also carried out unpolarized neutron scattering experiments, to take advantage of the increased intensity available.
Knowing, from the polarization analysis experiments, that the nuclear scattering is essentially constant over the temperature range 8K-295K, and that the magnetic scattering decreases to near zero at a temperature just above room temperature, a subtraction of the unpolarized scattering cross section at 295K from the cross section at 8K yields the magnetic scattering cross section at low temperatures. The results of this procedure are shown in Fig, 5. The splitting of the magnetic diffuse peaks is observed to increase as the Mn concentration decreases. It is seen that the absolute magnitudes of the magnetic differential scattering cross sections (as obtained with a calibration to the incoherent scattering cross section of vanadium) for the polarized and unpolarized data are in reasonable agreement.

DISCUSSION
It is clear that the magnetic diffuse scattering is symmetry-related to the short range order nuclear scattering. We interpret the scattering along the line [K.,1,0] to be due to one of three possible orientations of short range ordered clusters. That is, the magnetic diffuse peaks at [h ± 6,1,0] along this line arises from scattering from the same regions of the macroscopic crystal as those that give rise to the diffuse nuclear 4.
In view of the fact that the small angle neutron scattering from Cu-Mn alloys clearly shows the freezing of spin fluctuations, (8) it may appear surprizing that this magnetic diffuse scattering displays no hint of a freezing transition at low temperatures. Hov/ever, we believe that this data is precisely the information we have needed to justify the qualitative model which has emerged over many years and has guided our thinking about magnetism in these spin glass alloys (13). This model involves the formation of magnetic clusters at some temperature well above T^, and an increase in their concentration and average moment as the temperature is lowered toward T-. The fluctuations within and among these clusters slows down as T approaches T-. Whether or not critical slowing down should be expected is not cTear. We would like to suggest that the model magnetic clusters are, in fact, the short range order domains, and that the data of figs. 2,3,4 and 5 represent the Fourier decomposition of the magnetism within these regions. With this data as the basis, we are currently working on various models to obtain a physical picture of the magnetic structure of these clusters in real space. We are also pursuing both inelastic and small angle neutron scattering experiments on these single crystal samples.