We show that NMR chemical shift spectroscopy could help to identify the crystalline phases of hard carbon nitride compounds. To this purpose we compute the NMR chemical shifts of defect zinc-blende, cubic, a2, b2, and graphitic C 3 N 4 with a newly developed ab initio method. The C shifts can be used to identify the CN bonds and to characterize C hybridization. The N shifts distinguish the a-C 3 N 4 from the b-C 3 N 4 phases, and indicate the presence of the graphitic phase. [S0031-9007 (98)05665-8] PACS numbers: 76.60.Cq, 71.15.Mb, 82.80.Ch Carbon nitride compounds have been predicted [1] to exhibit extraordinary hardness, with some structures possibly being harder than diamond [2]. This has stimulated substantial research efforts in search of these materials, employing both experimental [3-14] and theoretical [2,13,15,16] methods.Shortly after the possibility of high bulk modulus due to the expected short C-N bond lengths was recognized [1], the b-C 3 N 4 [2] and later the defect zinc-blende [16], a-C 3 N 4 [2], and cubic-C 3 N 4 [2] structures were proposed. The theoretical bulk moduli of these structures are 427, 425, 425, and 496 GPa, respectively. Experimental studies, mostly on thin films, followed soon, finding nanometer to micrometer sized crystals embedded usually in an amorphous matrix [3,6,8,11,12]. In several cases [3,6,11] the crystallites were believed to be of the b-C 3 N 4 structure, but it has been pointed out later [2] that the a-C 3 N 4 structure would comply better with the experimental data. Crystallites with a tetragonal structure [12] and even a monoclinic phase [12] have been reported recently. When using a chemical precursor technique, the defect zincblende structure also appears to form [13].In all experiments, the crystallites obtained are too small to allow a conclusive x-ray diffraction (XRD) analysis. Therefore, structural determinations have been done with XRD in combination with other techniques, e.g., electron diffraction [3,6,8,[11][12][13]], x-ray photoelectron spectroscopy [3,8,10], electron energy loss spectroscopy [4], backscattering [3,6,8,10,11,17], and transmission electron microscopy [6,11,12,14]. Still, the determination of the structure of the crystallites remains difficult and not definitive, often because of the obscuring signals from the amorphous matrix or the substrate [11].In this Letter, we suggest the use of nuclear magnetic resonance (NMR) chemical shift for C-N compound sample characterizations. NMR shift experiments are becoming increasingly popular for characterizing thin-film samples of amorphous carbon [18][19][20][21][22][23] and silicon [24]. By applying a recently developed ab initio method [25], we predict the NMR chemical shifts for C and N atoms of five different low-energy C 3 N 4 structures. Our theoretical data allow an interpretation of experimental NMR spectra, thereby opening a new way for C-N compound characterization. Unlike XRD, NMR experiments do not require long range order, and, as we shall see, should be able to clearly identify the sig...