In perovskite-type doped manganites, the phase separation phenomenon is considered to be a probable candidate for the origin of the colossal magnetoresistance effect. This is because the magnetresistance effect is enhanced in the vicinity of the ferromagnetic metallic — charge-ordered insulating phase boundary, where the phase separation effect is also enhanced. Machida et al., however, experimentally indicated that the phase separation of Nd0:55(Sr,Ca)0:45MnO3 takes place at higher temperature ( 200K) than the insulator-metal transition temperature at 150K, and the mass fraction of the insulating phase monotonously increases with further decreases of temperature. This observation indicates that the phase separation has no direct connection with the insulator-metal behavior. Then, the remaining problem is the origin for the lowtemperature phase separation of the perovskite manganites. In this paper, we briefly report structural properties of single-layered manganites, Nd1 xSr1þxMnO4 (0:3 x 0:95). This compound begins to attract current interest due to the rich phase diagram. We found that the compound consists of two single-layered phases even at room temperature in the concentration range of 0:75 x 0:95, probably due to the spinodal decomposition with the Sr2þ concentration as the order parameter. We will compare the present spinodal decomposition in Nd1 xSr1þxMnO4 with the phase separation observed in the perovskite manganites. The polycrystalline samples were synthesized by solidstate reaction in air. A stoichiometric mixture of commercial Nd2O3, SrCO3, and Mn3O4 powders was well ground and calcined twice at 1350 C for 24 h. Then, the resulting powder was pressed into a disk with a size of 20mm 5mm and sintered at 1350 C for 48 h. The temperature of the sintered pellet was cooled down very slowly ( 50K/h). The neutron powder diffraction profiles were obtained at room temperature, using the Kinken powder diffractometer for high efficiency and high resolution measurements (HERMES) installed at the JRR-3M reactor at the Japan Atomic Energy Research Institute, Tokai, Japan. Neutrons with wavelength 1.8196 A were obtained by the 331 reflection of the Ge monochromator, and 120-B-Sample-220 collimation. The sample pellet was crushed into fine powders and sealed in a vanadium capsule (10mm ). The diffraction data were integrated for 3 h. We have performed Rietveld analysis on the obtained powder patterns with use of the RIETAN2000 program. Figure 1 shows prototypical examples of neutron powder pattern (cross) of Nd1 xSr1þxMnO4 at 300K, together with the Rietveld refinement (solid curve): (a) x 1⁄4 0:6 and (b) x 1⁄4 0:8. At x 1⁄4 0:6 ((a)), the tetragonal model (I4=mmm; Z 1⁄4 2) well reproduced the diffraction patterns in the range of 5 2 150 . The final refinements are satisfactory, in which Rwp and RI (reliable factor based on the integrated intensity) are fairly reduced. This model, however, fails to reproduce the powder patterns for 0:75 x 0:95. We tried several structural models, and finally found that the twophase model with two tetragonal (I4=mmm; Z 1⁄4 2) structures satisfactory reproduces the patterns, as exemplified in Fig. 1(b). Hereafter, we distinguish the two single-layered phases by the lattice constant c, and call then as the ‘long-c’ and ‘short-c’ phases for convenience of the explanation. We summarized in Fig. 2 thus obtained structural parameters of Nd1 xSr1þxMnO4: (a) lattice constants, a and c, (b) Mn–O bondlength d and (c) volume fraction s of the long-c phase. Circles and squares represent the short-c and long-c phases, respectively. With increase of x, c (open symbols) steeply decreases, and then gradually increases above x 0:6. On the other hand, a (filled symbols) remains nearly constant up to x 0:7. These x-dependent features of the lattice constants are similar to the case of La2 2xSr1þ2xMn2O7. 7) In 0:75 x 0:95, the compound consists of two single-layered phases with different lattice constants [see Fig. 2(a)]. In a thermodynamical point of view, the two-phase feature means that Nd1 xSr1þxMnO4 is unstable in 0:75 x 0:95, and decomposes into the Sr2þpoor (short-c phase) and Sr2þ-rich (long-c phase) phases (spinodal decomposition). The bottom panel [Fig. 2(c)] shows the mass fraction s of the long-c phase. s monotonously increases as x increases. The middle panel [Fig. 2(b)] shows Mn–O bondlength d 40 60 80 100 0 5 10