The rare-earth tetraborides, RB4 with a tetragonal structure of space group P4=mbm (No. 127), have been attracting interest from the viewpoint of geometrical frustration. The lattice of rare-earth ion consists of squares and triangles, whose configuration within the c-plane is topologically equivalent to the two dimensional model studied by Shastry and Sutherland. Since the magnetic interaction in RB4, probably of Ruderman–Kittel–Kasuya–Yoshida (RKKY) type, is three dimensional and long-ranged, most compounds exhibit long-range magnetic orderings; therefore, the frustration effect may not be so apparent. However, various kinds of anomalous properties have been revealed recently, which we speculate to be relevant to the frustrated lattice. TbB4 exhibits two phase transitions at TN1 1⁄4 44K and TN2 1⁄4 24K. Magnetic susceptibilities for H k a ( a) and H k c ( c) both show a cusp at TN1, followed by a weak kink at TN2. 4) Although the cusp at TN1 for both field directions suggests existence of c-axis and c-plane components in the antiferromagnetic moments, the magnetic structure deduced from neutron diffraction is collinear type, where the moments are orientated along the a-axis and involves no c-axis component. However, the magnetic structures of GdB4, 7,8) DyB4, 9) and HoB4, 10) which have been clarified recently, are non-collinear and the latter two have c-axis component as well. Therefore, it is necessary to reinvestigate the magnetic structure of TbB4. The difference in the structure between the intermediate phase (TN1 T TN2) and the lowtemperature phase (TN2 T) has not yet been clarified. In addition, recent discovery of the multistep magnetization process in high fields for H k c also requires detailed knowledge on the magnetic structure of TbB4. 11) Powder sample of TbB4 was obtained by reacting terbium oxide and boron powders in vacuum using a high frequency furnace. The reaction was repeated twice to reduce the impurity phase of TbB6. B isotope was used to prevent absorption of neutrons by B. The magnetic susceptibility of the powder sample showed a cusp at TN1 1⁄4 44K and a kink at TN2 1⁄4 22K. Neutron powder diffraction experiment was performed on the Kinken powder diffractometer for high efficiency and high resolution measurements, HERMES, of Institute for Materials Research, Tohoku University, installed at the JRR-3 reactor in JAEA, Tokai. Neutrons with a wavelength of 1.8142 A were obtained by the 331 reflection of Ge monochromator and the collimation of 120-blank-sample-220 was employed. The powder sample of 5.053 g was filled in a vanadium cylinder with a diameter of 7mm and was sealed in a standard aluminium cell with helium gas. The cell was cooled with a closed cycle He-gas refrigerator. Figure 1 shows the powder pattern for the three phases of TbB4. The pattern at 50K shows the nuclear peaks, impurity peaks of TbB6, and relatively large background with a significant modulation around 2 13 . This modulation disappears in the pattern at 300K, indicating that this is due to magnetic diffuse scattering. As demonstrated in the inset, the intensity decreases abruptly below TN1. Note that the thermal diffuse scattering also contributes to the background at high temperatures, especially to the data at 300K. Although it is an interesting viewpoint whether the diffuse scattering is relevant to the frustration, there seems no significant difference in the intensity just above TN as compared with that of TbB2C2 without geometrical frustration. Structural parameters at 300K were refined by the Rietveld method with RIETAN-2000. The parameters obtained are as follows: a 1⁄4 7:120ð1Þ A, c 1⁄4 4:0294ð7Þ A, xTbð4gÞ 1⁄4 0:3172ð2Þ, zB1ð4eÞ 1⁄4 0:2007ð6Þ, xB2ð4hÞ 1⁄4 0:0869ð2Þ, xB3ð8jÞ 1⁄4 0:1773ð2Þ, yB3ð8jÞ 1⁄4 0:0395ð2Þ. Fraction of TbB6 impurity phase is 7.46%, and the reliability factors are Rwp 1⁄4 5:85, Re 1⁄4 3:34, and S 1⁄4 1:75. In order to determine the magnetic structure, we compared the calculated powder patterns for possible models with the experimental data. Since all the magnetic peaks in Figs. 1(a) and 1(b) can be indexed by integer numbers, we see that the unit cell of the magnetic structure is the same as that of the chemical one. Note that the magnetic peaks of TbB6 impurity phase, with TN 1⁄4 19:5K, is recognized in Fig. 1(a), which can be indexed by q 1⁄4 ð1=4; 1=4; 1=2Þ. There are twelve sets of basis structures that are classified by the irreducible representations; three of them are picked up in Fig. 2. Model B is realized in GdB4, model A in the intermediate phase of DyB4, and superposition of models A and C in the lowest temperature phase of DyB4 and HoB4. 7–10) We calculated the powder pattern for the twelve models by considering the nuclear and magnetic structure factors, Lorentz factor, scale factor, and absorption correction. We used the result of Rietveld analysis at 50K to calculate the nuclear structure factor. Gaussian peak profile was assumed, whose width was obtained from the fitting at 50K. Background was fitted with a polynomial function. Calculated pattern for 50K is shown by the solid line in Fig. 1(c). At 25K, model B in Fig. 2 best explains the data. As shown by the solid line in Fig. 1(b), the intensities of all the reflections can be reproduced by the calculation, which is not the case in other models. Therefore, in the region TN1 T TN2, the magnetic structure is expressed by the model B, where the moments lie in the c-plane and are noncollinear. The magnitude of the magnetic moment at 25K was deduced to be 7.2 B from the scale factors of nuclear and magnetic reflections. The powder pattern at 3K can be explained well only by increasing the magnetic scale factor on model B. However, there are a few reflections whose intensities cannot be reproduced. The most typical case is the 110 reflection. Although the magnetic structure factor vanishes in model B, the intensity clearly increases below TN2 as shown in Fig. 3(a). This increase can be explained by mixing the model A or C as demonstrated by the solid line. The B+A E-mail: tmatsu@iiyo.phys.tohoku.ac.jp Journal of the Physical Society of Japan Vol. 76, No. 1, January, 2007, 015001 #2007 The Physical Society of Japan