Optical signature of strong electron correlations in transition metal monosilicides

(F. P. Mena)

Among the transition metal silicides, three of them have particularly attracted the attention of the scientific community for more than five decades. These are MnSi, FeSi and CoSi. All of them crystallize in the same cubic structure, B20, whose main property is the lack of strict space-inversion symmetry. In MnSi, this leads to the helical ordering in its magnetic phase. The Curie temperature (T_C) of this material is 30 K at ambient pressure but it can be decreased by the application of hydrostatic pressure and eventually becomes zero at p_c=14.6 kbar [C. Pfleiderer et al. Nature 414, 427 (2001)]. On the other hand, FeSi is paramagnetic down to at least 0.04 K [M. B. Hunt et al., Phys. Rev. B 50, 14933 (1994).] and CoSi is diamagnetic with a temperature independent susceptibility [J. H. Wernick et al., Mat. Res. Bull. 7, 476 (1972)]. The solid solutions between them, Fe_1-xCo_xSi, also crystallize in the B20 structure with the metal sites being occupied randomly by Fe and Co atoms. However, in contrast to their parent compounds, they show helimagmetism [J. Beille et al., J. Phys. F: Metal Phys. 11, 2153 (1981)] with low Curie temperatures in the concentration range 0.05 < x < 0.8 (Fig. 1).

MnSi

Usually, manganese develops strong magnetic moments whether it forms molecules (e.g. Mn17) or is part of a compound (e.g. manganites). In MnSi, however, despite having a susceptibility that follows a Curie-Weiss behavior compatible with a magnetic moment of 2.5 uB in the non-magnetic phase (above T_C=30K), its magnetic moment below T_C, extracted from the saturation magnetization, it is of just 0.4 uB. It has been sugested therefore, that in MnSi, manganese does not form local moments and instead, the magnetic properties are better explained in an itinerant picture where the magnetic moments are not localized in real space but in k-space. The problem with this picture is that usually applies to compounds whose original constituents do not present magnetism in its elemental form. Experimentally, we have discovered that several physical quantities do not follow the predictions of the itinerant picture. For example, the resistivity just above T_C is not proportional to T^5/3 (as seen in, e.g., ScIn₃). Other quantity which seems to be in odds with this picture is the optical conductivity where we have seen an strong mass enhancement and a frequency-dependent scattering rate which  is not proportional to w² (Fig. 2). This latter property is even more intriguing if we take into account that the resistivity below T_C is proportional to T² (since energy and temperature are equivalent, the same dependence is expected in both cases). Other surprises might be waiting to be discovered in this system and at the present we have plans to determine the optical Kerr response and get insight in the way the magnetic moment develops.

Figure 1: Phase diagram of the silicides from MnSi to FeSi and to CoSi [adapted from Mamyala et al., Nature 404, 581 (2000)]. The arrows indicate the materials of which the optical properties were determined.	Figure 2: Frequency dependent mass enhancement and scattering rate [from F. P. Mena et al., Phys. Rev. B 67, 241101 (2003)].

FeSi

Despite the fact that band calculations describe reasonably well the ground state properties of FeSi, there are several properties that cannot be accounted for. It has been suggested that this is the result of FeSi being a Kondo insulator (Kondo insulators can be seen as strongly correlated insulators in the same way that heavy fermion systems can be seen as strongly correlated metals). Tough several experiments point to this situation, this is still matter of controversy. The most direct way of determining whether or not strong correlations are important, is by studying the spectral weight contained in the optical conductivity. The spectral weight is just the integrated area of the optical conductivity. In normal insulators (i.e. ones in which all its properties can be explained from the band structure) the spectral weight is conserved when going from the insulating state to when the gap is thermally filled. This recovery takes place in the energy region just above the gap. In contrast, in Kondo insulators, when the gap is filled, the spectral weight recovers in an energy range much larger than the gap itself. This is demostrated in Fig. 3 where we have plotted the optical conductivity normalized to the one at 300 K. Some recovery takes place from 3000 to 4000 cm-1 (dark red) but it is not enough to recover completely the spectral weight lose with the opening of the gap (blue).

Figure 3: Contour plot of the optical conductivity of FeSi at different temperatures normalized to that one at 300 K.

Fe_1-xCo_xSi

The replacement of Co for Fe closes the gap present in FeSi. However, for x=0.1, 0.2 there is also a diminution of the spectral weight at low temperatures which is not recovered  in our spectral window. This implies that strong electron correlations are still presents in this material. Furthermore, in the x = 0.2 and 0.3 samples, we have seen an additional decrease of the spectral weight at low frequencies related with the transition to the magnetic state. This is opposite with what occurs in other magnetic compounds as manganites, EuB₆ or Ga_1-xMn_xAs. A possible explanation is the in the density of states as a result of Altshuler-Aranov effect which has been cited to explain the singular transport measurements in this system. What is surprising, it is that the Alshuler-Aranov effects appears usually just at extremely low temperatures in paramagnetic materials.