Вы находитесь на странице: 1из 6

Electronic Origin of Elastic Properties of Titanium

Carbonitride Alloys
V.I. IVASHCHENKO, P.E.A. TURCHI, A. GONIS, L.A. IVASHCHENKO, and P.L. SKRYNSKII

We have carried out numerical ab initio calculations of the elastic constants for several cubic ordered
structures modeling titanium carbonitride (TiCxN1x) alloys. The calculations were performed using
the full-potential linear augmented plane-wave method (FPLAPW) to calculate the total energy as
functions of volume and strain, after which the data were fit to the traditional Murnaghan equation of
state and to a polynomial function of strain to determine the formation energy; the bulk modulus; and
the elastic constants C11, C12, and C44. The predicted equilibrium lattice parameters are slightly
higher than those found experimentally (on average by 0.2 pct). The computed formation energy
indicates that the alloys are stable in the entire range of the carbon concentration x and the maximum
stability is obtained for 0.5 # x # 0.75. The computed bulk modulus, the shear modulus G, and the
Young’s modulus E are within approximately 2, 1, and 2 pct of the experimentally measured char-
acteristics, respectively. The maximum deviation is observed for TiC and TiN. The moduli G, E, and
Poisson’s ratio reach a maximum value at approximately the middle of the concentration range,
which is due to the fact that the shear modulus C44 shows a maximum value for a valence electron
concentration (VEC) in the range of 8.25 to 8.5. The other shear modulus (C11 – C12)/2 does not
exhibit any maximum overall concentration range and instead has a flat dependence in the range
mentioned previously. Such a concentration behavior of the elastic constants is related to specific
changes in the band structure of TiCxN1x alloys caused by the orthorhombic and monoclinic strains
that determine the shear moduli (C11 – C12)/2 and C44, respectively.

I. INTRODUCTION xBTiC 1 (1  x)BTiN. (4) The shear modulus (G) has a


slight maximum of 196 GPa for a VEC ; 8.4 to 8.3.[4]
TITANIUM carbonitride alloys (TiCxN1x) have To our knowledge, there have been relatively few theo-
attracted considerable interest due to their high hardness
retical studies on the mechanical properties of titanium
and melting temperature and comparatively high electrical
carbonitride alloys.[8–10] Using a muffin-tin orbital method,
and thermal conductivities, and are widely used for cutting
Zhukov et al.[8] revealed that the maximum value of the
tools and wear-resistant coatings.[1,2] The mechanical prop-
formation energy of TiCxN1x occurs at x 5 0.25 with the
erties of these alloys have been thoroughly investigated
magnitude of ;13 mRy, while the pseudo-potential total-
during the last 10 years. The review of such investigations
energy calculations[9] show that these characteristics reach
is done in References 1 through 4. In short, the following
a maximum value at x 5 0.75 and are approximately 4
main experimental findings were obtained. (1) The lattice
times lower. Jhi and Ihm[9] also found a positive deviation
parameter of the alloys is a linear function of the compo-
from the mixing rule for most of the different structure
sition between the carbide and nitride.[3] (2) There is no
types considered. Using an ab initio pseudo-potential
ambiguity in determining micro- and nanohardness (H) of
approach, Jhi et al.[10] studied the behavior of the shear
TiCxN1x as functions of composition. Results on well-
modulus C44 as a function of composition to interpret hard-
characterized specimens[3,4] show that H increases monot-
ness enhancement in TiCxN1x alloys.
onously with x, while the samples[5] exhibit a minimum at a
By comparing these findings, one can deduce that there is
valence electron concentration per cell (VEC) of about 8.5.
no agreement between them. The contradictions are ob-
On the contrary, the authors[1,6,7] report a maximum hard-
served between the experimental results as well as between
ness for the alloys in the middle of the composition range
the data obtained from various band structure calculations.
(VEC ; 8.4). (3) The bulk modulus (B) of TiCxN1x shows
Therefore, in the present investigation, we use another band
a slight negative deviation from the mixing rule B(x) 5
structure procedure to account for the elastic properties of
TiCxN1x alloys. For this purpose, ab initio self-consistent
full-potential linearized augmented-plane-wave (FLAPW)
V.I. IVASHCHENKO, Head of Laboratory, L.A. IVASHCHENKO, band structure calculations of several cubic structures
Senior Research Scientist, and P.L. SKRYNSKII, Research Scientist, are representing titanium carbonitride alloys have been per-
with the Institute of Problems of Material Science, NAS of Ukraine, 03142
Kyiv Ukraine. Contact e-mail: ivash@materials.kiev.ua P.E.A. TURCHI formed. The composition dependence of the lattice param-
and A. GONIS, Senior Research Scientists, are with the Lawrence Liver- eters (a); formation energy (EForm); elastic constants C11,
more National Laboratory (L-353), Livermore, CA 94551. C12, and C44; and the derived moduli B, G, and Young’s
This article is based on a presentation made in the symposium entitled modulus (E) have been calculated and thoroughly exam-
‘‘Fourth International Alloy Conference,’’ which occurred in Kos, Greece,
from June 26 to July 1, 2005, and was sponsored by Engineering Confer-
ined. The established trends are explained in terms of the
ences International (ECI) and co-sponsored by Lawrence Livermore peculiarities of the band structure changes under corre-
National Laboratory and Naval Research Laboratory, United Kingdom. sponding strains.

METALLURGICAL AND MATERIALS TRANSACTIONS A U.S. GOVERNMENT WORK VOLUME 37A, DECEMBER 2006—3391
NOT PROTECTED BY U.S. COPYRIGHT
II. COMPUTATIONAL ASPECTS 1/2C44V. Thus, once B, b(e), and b(g) are determined, one
can calculate the elastic moduli C11, C12, and C44. The use
For the present ab initio investigation, nonrelativistic of strains [1] and [2] enabled us to avoid the consideration
band structure calculations within the local-density approx- of the tetragonal and trigonal distortions that are sometimes
imation of density-functional theory were carried out using used.[17]
the FLAPW method[11] for the cubic supercells of Ti4C4, For the maximum values of the strains e and g, we use
Ti4C3N1, Ti4C2N2, Ti4C1N3, and Ti4N4 representing tita- the recommendation of Mehl.[14] In particular, we have
nium carbonitride alloys, TiCxN1x. Atomic relaxations chosen ei and gi, i 5 1,. . ., 5, and e5 5 0.05 and g 5 5 0.176.
were neglected, since they were found to be very small Because of the symmetry, the composition Ti4C2N2 was
with little impact on the behavior of the mechanical proper- calculated under the strains e and g for three atomic con-
ties as functions of composition.[9] The semicore states figurations in the nonmetal sublattice consisting of alternat-
were considered as valence states, while the core states ing the same layers in three directions. Therefore, the
were treated as atomic-like states. The wavefunctions were resulting values of the elastic moduli for this composition
expanded into a set of augmented plane waves with wave- were evaluated as the average values between these config-
vectors up to |k|Rs 5 8.0, where Rs is the smallest radius of urations.
the muffin-tin spheres. Inside these spheres, the potentials In the case of TiCxN1x alloys, obtained either by hot
and the charge densities were expanded in spherical har- pressing[4] or by using thin film technology,[18] experiments
monics up to l 5 8. In the interstitial region, the corre- can only determine the isotropic bulk modulus B and the
sponding expansions were carried out in Fourier series up shear moduli of polycrystalline aggregates of small crystal-
to |K| 5 14.0. The criterion of convergence for the total lites.[1–4,18] Then, the following expressions for determining
energy was 0.1 mRy/formula unit. The generalized gradient G, E, and Poisson’s ratio (s) are applicable[14,19]:
approximation for the exchange-correlation potential[12]
was employed. G ¼ (G1 1 G2 )=2
For the k-integration, the modified tetrahedron
G1 ¼ G1 1 3(G2  G1 )=[5  4b1 (G2  G1 )] [3]
method[13] was used throughout. In solving the Schrödinger
equation, different numbers of k-points were chosen G2 ¼ G2 1 2(G1  G2 )=[5  6b2 (G1  G2 )]
depending on the structure under investigation. We selected
the minimal sets of k-points (Nk) in the irreducible wedge where
of the Brillouin zone that guarantee the changes in the total
G1 ¼ (C11  C12 )=2, G2 ¼ C44
energy of about 1 mRy/formula unit by further increasing
the number of wave vectors. The following structures were b1 ¼ 3(B 1 2G1 )=[5G1 (3B 1 4G1 )],
considered: fcc (2 atoms/cell, Nk 5 165), SC (8 atoms/cell, b2 5  3(B 1 2G2 )=[5G2 (3B 1 4G2 )]
Nk 5 64), orthorhombic (2 atoms/cell, Nk 5 171; 8 atoms/
cell, Nk 5 64), and monoclinic (8 atoms/cell, Nk = 80). No
extrapolation to an infinite number Nk was undertaken. Associated with the bulk and shear moduli are the
The elastic moduli of a cubic crystal may be divided into Young’s modulus
two classes: the bulk modulus B 5 (C11 1 2C12)/3 and the
two shear moduli, (C11  C12)/2 and C44 (14). The bulk E ¼ 9BG=ð3B 1 GÞ [4]
modulus was calculated by fitting the total energy-volume,
ET(V), curve to the traditional Murnaghan equation of and the Poisson’s ratio
state.[15] To calculate the modulus (C11  C12)/2, we used s ¼ 1=2  1=6 E=B [5]
the volume-conserving orthorhombic strain[16]:
We have used these expressions to estimate the elastic
x0 5 (1 1 e)x constants of the alloys.
y0 5 (1 1 e)1 y [1]
0
z 5z III. RESULTS AND DISCUSSION
For the elastic modulus C44, we used the volume-con- The lattice parameters of TiCxN1x alloys are presented
serving monoclinic strain[16]: in Figure 1. One can see that the lattice parameter varies
almost linearly with composition. The calculated values
x0 5 x 1 gy slightly exceed the experimental ones[3] (by approximately
0.16 and 0.24 pct for TiC and TiN, respectively).
y0 5 y [2]
The results of the calculations of the formation energy,
z0 5 z presented in Figure 2, indicate that, in the entire concen-
tration range, it is energetically favorable for TiC and TiN
It should be noted that these strains differ from the strain
to mix and form alloys, in agreement with experiment.[8]
tensors suggested by Mehl.[14] The strains change the total
The polynomial fit to the calculated energies clearly points
energy as follows[14]:
to a minimum in the formation energy located in the com-
ET ðV,dÞ 5 a 1 bd2 1 cd4 1 O[d6 ] position range 0.5 , x , 0.75. The maximum absolute
value of the formation energy at T 5 0 K is 3.1 mRy, which
where V is the volume; a, b, and c are the fitting coefficients; is consistent with that obtained by Jhi and Ihm[9] (3.5
d 5 {e, g}; b(e/ 0) 5 (C11  C12)V; and b(g / 0) 5 mRy at x 5 0.75). The cohesive energies of TiC and TiN

3392—VOLUME 37A, DECEMBER 2006 METALLURGICAL AND MATERIALS TRANSACTIONS A


Table 1. Elastic Moduli (GPa) of TiCxN1x

Composition X C11 C12 C44 References


Ti4N4 0.00 583 123 155 a
636 158 — b
610 100 168 c
625 165 163 d
Ti4C1N3 0.25 556 124 175 a
Ti4C2N2 0.50 543 119 188 a
Ti4C3N1 0.75 533 112 184 a
1.00 516 112 166 a
513 130 — b
470 97 167 c
Ti4C4 1.00 513 106 178 d
540 110 180 e
515 106 179 f
500 113 175 g
a
Present investigation.
b
Results of the pseudo-potential calculations.[9]
c
Fig. 1—Lattice parameter of TiCxN1x vs composition. The experimental Elastic constants calculated by using the FPLMTO method with the GGA
data are the open circles.[8] approximation for the exchange and correlation potential.[21]
d
Experimental values presented in Ref. [22]
e
Results derived from neutron investigations.[23]
f,g
Results obtained from ultrasonic measurements.[24.25]

Fig. 2—Formation energy of TiCxN1x vs composition. Here and in the


following figures the solid line is the polynomial fitting to the data points
to be considered as a guide to the eye.

calculated in this work are 1.141 and 1.073 Ry, respec-


tively, compared with the experimental values of 1.04 and
0.97.[20] For the sake of comparison, the corresponding val-
ues obtained by using the pseudo-potential approach[9] were
1.292 and 1.216 Ry.
In Table I, we summarize the results of the calculations Fig. 3—Shear moduli (a) (C11 – C12)/2 and (b) C44 as functions of the
of the elastic moduli. For TiC, we note that our theoretical VEC. The open circles are the theoretical data from Ref. 10.
values of C11 and C12 agree well with the experimental
data,[22–25] while the calculated value for C44 is lower than
the experimental values. For TiN, all our elastic constants reaches the maximum around a VEC in the range 8.25 to
are lower than the experimental moduli. The maximum 8.5. The behavior of another shear modulus has nonmono-
deviation is observed for C12 (about 24 pct). As shown in tonic character. We see that the variation of the shear mod-
Table I, our results are consistent with those obtained in ulus (C11 – C12)/2 with VEC displayed a flat region around
previous theoretical investigations,[9,21] although the pseudo- a VEC of about 8.5.
potential moduli seem to agree better with the experimental To understand the variation of both shear moduli with
ones. For TiC, the FPLMTO approach underestimates C12, composition, let us analyze the changes in the structure of
while the pseudo-potential procedure strongly overesti- the energy bands and the density of states (DOS) of TiN
mates this modulus as compared to the experimental values. caused by the orthorhombic and monoclinic strains. The
The dependences of the shear moduli (C11 – C12)/2 and changes in the band-structure characteristics at other com-
C44 on the VEC are shown in Figure 3. For comparison, the positions are similar to those for TiN. The results of the
calculated values of C44[10] are also presented. It is seen that comparisons between the unstrained and strained band
both calculations give close values of this modulus that structures are shown in Figures 4 and 5. The orthorhombic

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 37A, DECEMBER 2006—3393


strain leads to an appreciable rise of the fourth and fifth bands TiC, the Fermi level lies below the top of the fourth band
relative to other bands along the symmetry lines G-D-X at the G point). When further increasing VEC, the positive
and X-Z-W, respectively, near the Fermi level (EF). As a contribution to the shear modulus can be associated with
result, the DOS in the region between the DOS minimum the fifth band along the line W-Z-X.
energy (Emin) and EF in the distorted material is lower than The extremal behavior of the C44 modulus as a function of
in the undistorted one (cf. Figure 5(a)). Consequently, when composition in TiCxN1x alloys was explained in Reference
raising the Fermi level from Emin toward higher energies, 10. The authors[10] have shown that the monoclinic strain
the positive contribution to the shear modulus (C11 – C12)/2 causes a rise of the fourth band along the G-S-K symmetry
related to the fourth and fifth bands will increase. This line and a lowering of the fifth band along this line. So, with
relation is confirmed by the results of the calculations of changing VEC, the fourth and fifth bands will make positive
the differences in the contributions to the total energy from and negative contributions to C44, respectively. Since this
the DOS of the distorted and undistorted TiN for energies assumption was not directly numerically confirmed, we cal-
higher than Emin (DE). In Figure 6(a), these differences are culated the DOS of the distorted and undistorted TiN and the
presented as a function of VEC. The bend in the DE(VEC) dependence DE(VEC) to clarify the influence of the band
dependence around VEC ; 8.3 to 8.4 corresponds to a structure on C44 in more detail. The results of the calcula-
modification of the band structure, when the fourth band tions are presented in Figures 5(b) and 6(b). One can
along the symmetry line G-D-X is completely filled. (In note that the evolution of the fourth band in TiN under a

Fig. 4—Electronic band structures of TiN at zero orthorhombic strain (left panel) and at finite orthorhombic strain (e 5 0.05) (right panel).

Fig. 5—Densities of states of TiN for (a) orthorhombic strain e 5 0 (solid line) and e 5 0.05 (dashed line) and (b) monoclinic strain g 5 0 (solid line) and
g 5 0.176 (dashed line). The vertical lines locate the Fermi levels associated with VEC 5 8 and VEC 5 9.

3394—VOLUME 37A, DECEMBER 2006 METALLURGICAL AND MATERIALS TRANSACTIONS A


Fig. 6—Differences between the band energies (DE) of TiN at the zero
and finite strains: (a) orthorhombic (e 5 0.05) and (b) monoclinic (g 5
0.176) as functions of VEC. The differences are relative to DE(VEC 5 8).
Fig. 7—Shear modulus (G), bulk modulus (B), and Young’s modulus (E)
for TiCxN1x vs composition. The open circles are the experimental data
monoclinic strain gives rise to an increase of the DOS below from Ref. 4.
Emin, and vice versa, the fifth band lowering leads to an
increase of the DOS above Emin. As a result, at small
VEC, DE (and C44) increases with VEC up to compositions
such that VEC ; 0.3. When further increasing VEC, the
filling of the fifth band gives rise to a decrease in DE (and
C44) (cf. Figure 6(b)). So, these findings are consistent with
the assumption made in Reference 10 and add new details in
the explanation of the extremal behavior of C44 as a function
of composition.
Finally, using the calculated elastic moduli B, C11, C12,
and C44, we estimated the elastic constants G (3), E (4), and
Poisson’s ratio s (5) that are experimentally measured.
These quantities as functions of composition are presented
in Figures 7 and 8. We can say that our results reproduce
fairly well the behavior of B and G with composition. For
B, we obtained a slightly less negative deviation from the
mixing rule determined experimentally.[4] Note that a pos-
itive deviation was found by Jhi and Ihm[9] for most struc-
ture types considered. For G, the FPLAPW approach Fig. 8—Poisson’s ratio (s) as a function of composition (full circles). The
slightly underestimates this modulus for TiC and TiN. This open circles are the values of s, determined from the experimental data of
shortcoming is reflected in the determination of the mod- Ref. 4 using expression [5].
ulus E. Our data show a distinct maximum at x ; 0.5 to
0.75, while the experimental values show only a gradual
change of E with composition (cf. Figure 7). We believe Poisson’s ratio derived from the experimental values of B
that this discrepancy can be resolved by further experi- and E[4] also speaks in favor of the later expectation
ments. The point is that the experiments[6,7] have revealed (cf. Figure 8). It is worth noting that the hardness is not
the greatest hardness in TiCxN1x for approximately the an easy-to-define physical quality. During an indentation
middle of the composition range, while the experiments test, a variety of material properties, such as dislocation
reported in Reference 4 indicate a gradual reduction of this motion and viscous material flow, can significantly influ-
characteristic on going from TiC to TiN. Therefore, one can ence the results. However, our theoretical investigations
expect that titanium carbonitrides with extremal composi- allow us to unambiguously deduce that in low defect
tional behavior of hardness will possess an extremal TiCxN1x alloys, one can expect an extremal concentration
dependence of the Young’s modulus on composition. The dependence of the E and G moduli and possibly hardness.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 37A, DECEMBER 2006—3395


IV. CONCLUSIONS 3. W. Lengauer, S. Binder, K. Aigner, P. Ettmayer, A. Guillou,
J. Debuigne, and G. Groboth: J. Alloys Compounds, 1995, vol. 217,
In conclusion, the application of the ab initio FPLAPW pp. 137-41.
procedure enabled us to calculate the formation energy; the 4. Q. Yang, W. Lengauer, T. Koch, M. Scheerer, and I. Smid: J. Alloys
Compounds, 2000, vol. 309, pp. L5-L9.
bulk modulus B; the elastic constants C11, C12, and C44; the 5. R. Kieffer, W. Wruß, K.K. Constant, and H. Haberman: Monatsh.
shear modulus G; the Young’s modulus E; and the Pois- Chem., 1975, vol. 106, pp. 1349-53.
son’s ratio s of titanium carbonitride alloys as functions of 6. H. Holleck: J. Vac. Sci. Technol., 1986, vol. A4, pp. 2661-69.
composition. The results revealed an extremal composition 7. V. Richter, A. Beger, J. Drobniewski, I. Endler, and E. Wolf: Mater.
dependence of the moduli G and E and Poisson’s ratio. Sci. Eng., 1996, vol. A209, pp. 353-57.
8. V.P. Zhukov, V.A. Gubanov, O. Jepsen, N.E. Christensen, and
Such a dependence is related to the fact that the shear O.K. Andersen: Phil. Mag., 1988, vol. B58, pp. 139-51.
modulus C44 shows a maximum value for a VEC in the 9. S.-H. Jhi and J. Ihm: Phys. Rev., 1997, vol. B56, pp. 13826-29.
range of 8.25 to 8.5. The other shear modulus (C11 – C12)/2 10. S.-H. Jhi, J. Ihm, S.G. Louie, and M.L. Cohen: Nature, 1999, vol. 399,
does not exhibit any maximum over the entire concentra- pp. 132-34.
11. P. Blaha, K. Schwartz, P. Sorantin, and S.B. Trickey: Comput. Phys.
tion range, and instead has a flat dependence in the afore- Commun., 1990, vol. 59, pp. 399-419.
mentioned range. Such a concentration behavior of the 12. J.P. Perdew and Y. Wang: Phys. Rev., 1992, vol. B45, pp. 13244-49.
elastic constants is related to specific changes in the band 13. P.E. Blöchl, O. Jepsen, and O.K. Anderson: Phys. Rev., 1994, vol. B49,
structure of TiCxN1x alloys caused by the orthorhombic pp. 16223-33.
and monoclinic strains that determine the shear moduli 14. M.J. Mehl: Phys. Rev., 1993, vol. B47, pp. 2493-500.
15. F.D. Murnaghan: Proc. Nat. Acad. Sci. U.S.A., 1944, vol. 30,
(C11 – C12)/2 and C44, respectively. Despite the contradic- pp. 244-47.
tory results on hardness of TiCxN1x alloys reported in the 16. A.O.E. Animalu: Intermediate Quantum Theory of Crystalline Solids,
literature, based on the present analysis of the effect of Prentice-Hall, Englewood Cliffs, NJ, 1977, pp. 132-81.
strains on electronic properties, we predict that hardness 17. M.J. Mehl, J.E. Osburn, D.A. Papaconstantopoulos, and B.M. Klein:
Phys. Rev., 1990, vol. B41, pp. 10311-23.
should exhibit an extremal concentration dependence. 18. A. Matthews and A. Leyland: Key Eng. Mater., 2002, vols. 206–213,
pp. 459-76.
19. Z. Hashin and S. Shtrikman: J. Mech. Phys. Solids, 1962, vol. 10, pp.
ACKNOWLEDGMENTS 335-39.
20. V.P. Zhukov, V.A. Gubanov, O. Jepsen, N.E. Christensen, and
This work was supported, in part, by Contract CRDF No. O.K. Andersen: J. Phys. Chem. Solids, 1988, vol. 49, pp. 841-60.
UK-E2-2589-KV-04. The work of two of the authors 21. R. Ahuja, O. Eriksson, J.M. Wills, and B. Johansson: Phys. Rev., 1996,
(PEAT and AG) was performed under the auspices of the vol. B53, pp. 3072-79.
United States Department of Energy by the University of 22. M.M. Choy, W.R. Cook, R.F.S. Hearmon, H. Jaffe, J. Jerphagon,
S.K. Kurtz, S.T. Liu, and D.F. Nelson: in Elastic, Piezoelectric,
California Lawrence Livermore National Laboratory under Pryelectric, Piezooptic, Electrooptic Constants and Nonlinear Dielec-
Contract No. W-7405-ENG-48. tric Susceptibilities of Crystals, K.-H. Hellwege and A.M. Hellwege,
eds., Landolt-Borstein, Numerical Data and Functional Relation-
ship in Science and Technology, Springer, Berlin, 1979, vol. 11,
REFERENCES pp. 280-93.
23. L. Pintschovius, W. Reichardt, and B. Scheere: J. Phys. C: Solid State
1. L.E. Toth: Transition Metal Carbides and Nitrides, Academic Press, Phys., 1978, vol. 11, pp. 1557-61.
New York, NY, 1971, pp. 90-153. 24. R. Chang and L.J. Graham: J. Appl. Phys., 1966, vol. 37, pp. 3778-83.
2. W. Lengauer: in Handbook of Ceramic Hard Materials, R. Riedel, ed., 25. I.J. Gilman and B.W. Roberts: J. Appl. Phys., 1961, vol. 32, pp. 1405-
Wiley-VCH, Weinheim, 2000, vol. 1, pp. 202-52. 06.

3396—VOLUME 37A, DECEMBER 2006 METALLURGICAL AND MATERIALS TRANSACTIONS A

Вам также может понравиться