Phys. 72, 283 (1988). 10. H. Granicher, Helv. Phys. Aca 29, 211 (1956).

5. D. Dimos et al., Phys. Rev. Lett. 61, 219 (1988). 11. We thank J. Berosh, D. R. Clarke, J. Hagerhorst, J.
6. J. Mannhart et al., Appl. Phys. Lett. 52, 1271 Lacey, M. M. Oprysko, and T. M. Shaw for support.
(1988). Part of this work was supported by a grant of the
7. R. Gross et al., IEEE Trans. Magn. 25, 2250 (1989). Bundesministerium fir Forschung und Technology
8. A. Barone and G. Patemo, Physics and Applications of (Projekt 13N5482).
the Josephson Effect (Wiley, New York, 1982).
9. A. T. Fiory et al., Phys. Rev. Lett. 61, 1419 (1988). 3 May 1989; accepted 20 June 1989

Prediction of New Low Compressibility Solids

AMY Y. Liu AND MARVIN L. COHEN
Fig. 2. Valence charge density of P-C3N4 in the
An empirical model and an ab initio calculation of the bulk moduli for covalent solids [0001] plane. The contours are in units of elec-
are used to suggest possible new hard materials. The empirical model indicates that trons per cell. The contour interval is 20 electrons
hypothetical covalent solids formed between carbon and nitrogen are good candidates per cell.
for extreme hardness. A prototype system is chosen and a first principles pseudopoten-
tial total energy calculation on the system is performed. The results are consistent with the range of 4.65 Mbar. In contrast, because
the empirical model and show that materials like the prototype can have bulk moduli both the bond length and ionicity are larger

Downloaded from https://www.science.org at National Taiwan University on April 14, 2024

comparable to or greater than diamond. It may be possible to synthesize such materials in BN than in diamond, Eq. 1 predicts that
in the laboratory. BN should have a smaller bulk modulus
than diamond. With X = 1 and d = 1.56 A,
HA ARDNESS (1) IS A COMPLEX PROP- ionicity of the compound. For the homopo- the scaling law yields B = 3.67 Mbar for
erty related to the extent to which lar semiconductors (group IV), A = 0, BN. A recent measurement (7) of the bulk
solids resist both elastic and plastic while for the heteropolar III-V and II-VI modulus of BN yields 3.69 Mbar. The good
deformation. For materials with defects, zinc blende solids, X = 1 and 2, respectively. quantitative agreement of the results of the
hardness can be limited by many factors Similar analytic relations between the bulk empirical model and experiment demon-
including point defects, dislocations, and modulus and lattice constant have been ob- strates the predictive power of the model.
macroscopic defects. On the microscopic tained for the group IV and III-V semicon- Based on this model, it was suggested (2)
level, for ideal systems, hardness is deter- ductors by means of first principles methods that a covalent solid formed between C and
mined by the bulk modulus, which in turn based on a pseudopotential scheme (4). N could have a larger bulk modulus than
depends on the nature ofthe chemical bond- The scaling law given by Eq. 1 successful- diamond. In this work, we investigate the
ing. It is the strength and compressibility of ly describes the diamond and zinc blende structural and electronic properties of a solid
the bond that plays the primary role in a semiconductors, giving results for the bulk of this type using first principles pseudopo-
solid's ability to resist deformation. The moduli within a few percent. This is compa- tential calculations (5). In choosing a candi-
largest bulk moduli are found in covalently rable in accuracy to first principles calcula- date structure that can be used as a proto-
bonded materials. Diamond, which has the tions (5) that require extensive supercom- type for a covalent C-N solid, the simple
largest bulk modulus (4.43 Mbar), is also puter time. Equation 1 is also useful for zinc blende structure is ruled out because in
the hardest known solid. A theory of the illustrating trends in these materials. For a hypothetical zinc blende compound the
bulk moduli of covalent solids is therefore example, to achieve a large bulk modulus, first antibonding band would be occupied.
useful for suggesting schemes to increase two properties are needed: low ionicity and This would distribute charge in antibonding
hardness. short bond length. In the past, there was regions and it is doubtful that such a struc-
An empirical model for the bulk moduli speculation that cubic BN might be harder ture would be stable. Instead, a more com-
of covalent solids was developed (2) with than diamond. Interpolation of empirical plex structure for a C-N solid, based on the
scaling arguments based on the Phillips- relations for the elastic constants (6) led to known P-Si3N4 structure with C substituted
Van Vechten scheme (3) for characterizing an estimate of the bulk modulus of BN in for Si, is considered. The octet rule for
the covalent and ionic nature of tetrahedral covalent bonding is satisfied in this struc-
solids by means of their spectral properties. ture, and no antibonding states are occu-
For purely covalent solids such as C and Si, pied. In Fig. 1, the structure is shown to
the bulk modulus scales as the average ho- consist of buckled layers stacked in an
mopolar energy gap in the reflectivity spec- AAA.. . sequence (8). The unit cell is hexag-
trum divided by the volume of the bond onal and contains two formula units (14
charge. This leads to the bulk modulus atoms) with local order such that C atoms
varying as d3 5, where d is the bond length. occupy slightly distorted tetrahedral sites
For the zinc blendes, an additional empirical while N atoms sit in nearly planar triply
term is added to account for the depletion of coordinated sites. This structure can be
bond charge with increasing ionicity. The thought of as a complex network of CN4
resulting scaling relation is Fig. 1. Structure of P-C3N4 in the a-b plane. The tetrahedra that are linked at the corners. The
c-axis is normal to the page. Half the atoms atomic coordination suggests sp3 hybrids on
B=(19.71 d3.5
- 2.20X) (1 illustrated are located in the z = -c/4 plane, the
(1) other half are in the z = c/4 plane. The structure
consists of these buckled planes stacked in Department of Physics, University of California, Berke-
where B is the bulk modulus in megabars, d ley, CA 94720, and Materials and Chemical Sciences
AAA ... sequence. The parallelogram shows the Division, Lawrence Berkeley Laboratory, Berkeley, CA
is in angstroms, and X is a measure of the unit cell. 94720.
25 AUGUST I989 REPORTS 841
the C atoms and Sp2 hybrids on the N atoms. the basis. The charge density and crystal the anion grows at the expense of the one
The hexagonal unit cell and the planar potential are calculated self-consistently with near the cation. In 13-C3N4, the two local
coordination of the N sites raise the possibil- an energy cutoff of 64 Ryd. maxima remain, but the charge is weighted
ity that this structure may exhibit anisotro- The structural properties of P-C3N4 are towards the N site. The height of the peaks
pic elastic properties. However, on a local determined by calculating the total crystal are 2.7 and 4.6. In BN, the charge is
scale, the CN4 tetrahedra are linked in such a energy as a function of volume. Since 1- strongly localized near the N site and the
way that the in-plane and out-of-plane C3N4 is used only as a prototypical struc- bond exhibits only a single peak of height
bonding are similar. The network is three ture, there is no reason a priori to believe 6.3. Thus the heteropolarity of the C-N
dimensional and is not obviously weak in that it is the lowest energy structure for the bond in 13-C3N4 lies intermediate between
any direction. This is supported by measure- C-N system. However, at the equilibrium purely covalent diamond and partially ionic
ments (9) of the elastic constants of 1-Si3N4 volume, the cohesive energy of 1-C3N4 is BN.
which show nearly equal in-plane and out- calculated to be 81 eV per cell or an average In conclusion, we suggest that it may be
of-plane linear compressibilities. Hence the value of 5.8 eV per atom. This moderately possible to synthesize a covalent C-N solid
1-Si3N4 structure is a good prototype for large cohesive energy suggests that there is a such as 3-C3N4 with compressibility compa-
this study even though the bonding is not good chance that 13-C3N4 is at least a meta- rable to diamond. A possible starting mate-
exclusively sp3. The lack of complete tetrahe- stable structure. rial is amorphous carbon nitride, which has
dral coordination suggests that Eq. 1 may At the calculated equilibrium volume, the a polymeric structure (14). The application
not be as precise as in the diamond and zinc average C-N bond length is found to be of high pressures and temperatures to cx-
blende cases, but it is expected that the 1.47A. This is in good agreement with the C: N may induce a phase transition to a
trends contained in the scaling law will be sum of the tetrahedral covalent radii (3, 12) structure with more tetrahedral bonding.

Downloaded from https://www.science.org at National Taiwan University on April 14, 2024

followed. Our ab initio study described be- of C and N. The bulk modulus of 13-C3N4 is This transition would be similar to the
low is intended to test this assumption. determined by fitting the total energy as a transition of graphite to diamond. Since
The structural and electronic properties of function of volume to Murnaghan's equa- other factors such as chemical reactivity and
the hypothetical compound 13-C3N4 (C3N4 tion of state (13) which yields a value of 4.27 fracture toughness are important for appli-
in the 13-Si3N4 structure) are calculated by (+0.15) Mbar. For the calculated bond cations, the prediction of hardness does not
means of the first principles pseudopotential length and an estimated ionicity factor of guarantee usefulness in engineering applica-
total energy approach (5). The local density X = 1/2 (discussed below), the scaling law tions. Nevertheless, the fabrication of a co-
approximation, with the Wigner interpola- (Eq. 1) overestimates the bulk modulus of valent solid formed between C and N could
tion formula (10) for the exchange and 13-C3N4 by about 10%. A similar overesti- result in an important new high-perform-
correlation functional, is used. Because there mate is found for the bulk modulus of 1- ance material that may best be used as part
are no p states in the core of C and N, the 2p Si3N4 which has a measured (9) value of 2.5 of a composite. From the scientific point of
pseudopotentials for these atoms are strong- Mbar. These overestimates are most likely view, the fabrication of a low compressibil-
ly attractive. Hence a localized orbital for- caused by the structural difference between ity C-N solid would add support for the
malism (11), in which the electronic wave these materials and the zinc-blende systems theoretical approach used here and provide
functions are expanded in linear combina- on which Eq. 1 was based. In particular, the information for further applications.
tions of Bloch sums of Gaussian orbitals trigonal coordination of N in f3-C3N4 and 1-
localized on the atomic sites, is employed. Si3N4 allows only three electrons per N site REFERENCES AND NOTES
Orbitals of s and p character are included in to participate in the bonding. The remain- 1. See, for example, Proceedings of the Third International
ing N electrons occupy nonbonding p orbi- Conference on the Science of Hard Materials, V. K. Sarin,
tals perpendicular to the bonding plane. Ed. (Elsevier, Oxford, 1988).
2. M. L. Cohen, Phys. Rev. B 32, 7988 (1985).
This reduction in the number of bonds 3. J. C. Phillips, Bonds and Bands in Semiconductors
6- ---- Diamond compared to a tetrahedral system results in (Academic Press, New York, 1973).
4. P. K. Lam et al., Phys. Rev. B 35, 9190 (1987).
elastic constants that are slightly lower than 5. M. L. Cohen, Phys. Scr. Ti, 5 (1982).
BN those predicted by Eq. 1. 6. J. A. Sanjuro, E. Lopez-Cruz, M. Cardona, Phys.
The degree of ionicity in P-C3N4 can be Rev. B 28, 4579 (1983).
7. E. Knittle et al., Nature 337, 349 (1989).
seen in plots of the charge density. Figure 2 8. 0. Borgen and H. M. Seip, Acta. Chem. Scand. 15,
shows the total valence charge density in the 1789 (1961).
9. L. Cartz and J. D. Jorgensen, J. Appi. Phys. 52, 236
C-
[0001] plane of P-C3N4. This plane contains (1981).
a central N atom surrounded by its three C 10. E. Wigner, Phys. Rev. 46, 1002 (1934).
neighbors. The bonds are highly directional, 11. J. R. Chelikowsky and S. G. Louie, Phys. Rev. B 29,
3470 (1984); C. T. Chan, D. Vanderbilt, S. G.
as expected for a covalent system. For com- Louie, ibid. 33, 2455 (1986).
parison, linear plots of the charge density 12. C. Kittel, Introduction of Solid State Physics (Wiley,
along a bond in diamond and BN, as well as New York, ed. 6, 1986).
13. F. D. Murnaghan, Proc. Nat. Acad. Sci. U.S.A. 30,
P-C3N4, are given in Fig. 3. To facilitate 244 (1944).
comparison between the different structures, 14. H.-X. Han and B. J. Feldman, Solid State Commun.
the charge densities have been normalized 65, 921 (1988).
15. Supported by National Science Foundation (NSF)
by the total number of electrons per cell. In grant DMR8818404 and by the Director, Office of
diamond, there are two local maxima along Energy Research, Office of Basic Energy Sciences,
Materials Sciences Division of the U.S. Department
C-N
the bond. The presence of the double hump of Energy under contract DE-AC03-76SF00098.
B N rather than a single maximum near the Cray computer time was provided by the U.S. Office
center of the bond is a result of the strongly of Energy Research of the Department of Energy at
Fig. 3. Normalized valence charge density along attractive C 2p potential. The two maxima
the National Magnetic Fusion Energy Computer
the bond in diamond, 13-C3N4, and BN. The Center and by the NSF at the San Diego Supercom-
normalization factor, p0, is the number of valence reach a normalized height of 2.9. With puter Center.
electrons per cell. increasing ionicity, the local maximum near 17 May 1989; accepted 23 June 1989

8,+2 SCIENCE, VOL. 24S