濃厚絶縁体スピングラスFe_xMn_<1-x>TiO_3の帯磁率測定による研究(修士論文(1985年度))

Abstract

Since the pioneering studies by Cannella and Mydosh, and by Edwards and Anderson, many experimental and theoretical investigations have been done on spin glasses which are formed in spin systems in which ferromagnetic and antiferromagnetic exchange interactions are randomly distributed. Experimentally the low field susceptibility shows a sharp cusp at a low temperature T_g, indicating that the local spins are frozen randomly in direction. On the other hand, the specific heal does not exhibit any evidence for a phase transition at T_g, and no long-range order has been observed by neutron scattering measurement. These features have been observed widely in dilute magnetic alloys in which the long-range RKKY interaction dominates and also in insulating compounds in which the nearest-neighbor exchange interaction does. The most fundamental question in properties of spin glasses is whether the spin glass freezing is a thermodynamic phase transition, or not. Many theoretical investigations have been done on Ising spin systems by using the mean-field approximation, on the assumption that the freezing is a phase transition. But the magnetic behavior of spin glasses experimentally examined does not always agree with that obtained by the mean-field approximation. The competition between exchange interactions in metallic spin glasses or diluted insulating ones is rather complex. In order to understand the physics of spin glasses, a non-diluted insulating spin glass is more suitable to be investigated, because the nature of the exchange interaction is well-known in it. But in such spin glasses ever examined, there exists not only the competition between the exchange interactions but also the other competition, such as that between the spin anisotropies. A non-diluted insulating compound Fe_xMn_<1-x>TiO_3 is expected to be an ideal Ising spin glass in which only the competition between the exchange interactions exists. Both FeTiO_3 and MnTiO_3 are antiferromagnets having easy-axis aisotropy along the hexagonal c-axis. Within the c-layer, spins are ferromagnetically coupled in FeTiO_3, and antiferromagneticaliy in MnTiO_3. In the mixture Fe_xMn_<1-x>TiO_3, therefore, the ferromagnetic and antiferromagnetic interactions compete with each other within the c-layer, and the spin glass behavior is expected to appear at low temperatures. Single crystals of Fe_xMn_<1-x>TiO_3 were investigated by magnetization and ac susceptibility measurements. As a result, in Fe_<0.50>Mn_<0.20>TiO_3 Ising-like spin glass behavior appears below T_g=21.1K, and we report experimental results about Fe_<0.50>Mn_<0.50>TiO_3. The zero-field cooled magnetization M_<ZFC>, which is measured in magnetic field of 20e along the c-axis(H//c) after zero-field cooling, shows a sharp cusp at T_g=21.1K. While the field cooled magnetization M_<FC> is nearly independent of temperature below T_g. Comparing the temperature dependence of magnetization measured in H//c with that done in H⊥c, Fe_<0.50>Mn<0.50>TO_3 is considered as an Ising-like spin glass. With increasing applied field H, the cusp is broadened, and the temperature T_g(H), where the irreversibility appears between M_<ZFC> and M_<FC>, decreases. The H dependence of T_g(H) is described by H = A[1 - T_g(H)/T_g(O)]^α = 4.26 × 10^4[1 - T_g(H)/T_g(O)]^<1.49> (1) This relation coinsides with the de Almeida-Thouiess line(α>1.5) calculated by using the mean-field approximation. Below T_g, the long-time relaxation phenomenon was observed. The time dependences of the thermoremanent magnetization M_<TRM> and the growth of the isothermal remanent magnetization M_<IRM> in applied field were investigated. It is found that M_<TRM> and M_<IRM> obey the power law quite well, M_<TRM>(t) = At^<-1> (2) M_<IRM>(t) = M_0 - At^<-n> (3) where M_0 is an adjustable parameter. Eq.(3) indicates that the value of M_<IRM> reaches M_0 after infinit time, that is, M_0 is the value of M_<IRM> in the thermodynamic equilibrium slate. The value of M_0 agrees with that of M_<FC> within experimental accuracy. This result suggests that the field cooled state is close to the thermodynamic equilibrium state. This conclusion is obtained, for the first time, from the relaxation measurement. The exponent n and the prefactor A of Eqs.(2) and (3), depend on both temperature and applied field. As temperatures increase, n increases exponentially, while A decreases linearly and is extrapolated to zero at near T_g(H). The temperature dependences of n and A for another spin glass Rb_2(Mn,Cr)Cl_4 for which the relaxation is fitted to the power law decay, are similar to those for Fe_<0.50>Mn_<0.5>TiO_3, though the authors have not touched upon them. Therefore, our findings about the temperature dependences of n and A are considered to be common at least for non-diluted insulating spin glasses. The ac susceptibility has been measured in a frequency range between 20Hz and 66.5kHz. The temperature T_g(f) showing cusp increases with increasing frequency f, which indicates that there exists wide distribution in relaxation times T=f^<-1>. The f dependence of T_g(f) is described by both the Fulcher law and the power law within this frequency range. Even if we take into account the result of the Mossbauer measurement ( T_g(f &sime; 10^7Hz) &sime; 28K), still both laws are equally well fitted. In order to determine which law is better, measurements by other technique with large f is required.