You are on page 1of 17

Acta Materialia 54 (2006) 1641–1657

www.actamat-journals.com

First principles determination of elastic constants and chemical bonding


of titanium boride (TiB) on the basis of density functional theory
K.B. Panda 1, K.S. Ravi Chandran *

Department of Metallurgical Engineering, University of Utah, 135 South 1460 East, Room 412, Salt Lake City, UT 84112, United States

Received 6 January 2005; received in revised form 25 November 2005; accepted 8 December 2005
Available online 27 January 2006

Abstract

First principles calculations of anisotropic elastic constants of titanium boride (TiB) have been performed using the computational
implementation of density functional theory (DFT). TiB has orthorhombic crystal structure, thus, nine independent elastic constants need
to be determined to completely characterize its polycrystalline elastic behavior as well as elastic anisotropy. TiB, especially in the whisker
form, has attracted attention recently as reinforcement in metal matrix composites, wear resistance coatings and has potential as mono-
lithic material due to its high hardness and elastic modulus coupled with its electrically conducting nature. In this study, the elastic con-
stants were determined using the WIEN2K computational implementation of the full-potential-linear-augmented-plane-wave (FLAPW)
method and the generalized gradient approximation (GGA). Nine independent elastic distortions of the unit cell were employed to deter-
mine the anisotropic elastic constants. Internal atomic relaxations after elastic distortions have been shown to have significant effects on
the numerical values of elastic constants. Polycrystalline elastic moduli were determined from the elastic constants and were compared
with that extrapolated from experimental data. The nature of chemical bonding and the electronic charge density distribution in TiB have
also been explored to explain the high hardness and high stiffness values as well as the nature of elastic anisotropy.
Ó 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: First-principle electron theory; Density functional; Full-potential linear augmented plane wave; Titanium boride; Elastic behavior

1. Introduction in situ as long and pristine whiskers, and its role as an effec-
tive stiffener of titanium matrix is very promising [5–8].
Titanium and boron react to form an interesting set of Recent research efforts [6,7,9–12] help to highlight the sev-
compounds, TiB, Ti3B4, and TiB2 at B concentrations eral property advantages that can be gained by incorporat-
(wt.%) of about 18%, 22%, and 30% [1] and they have scien- ing the high modulus (estimated to be about 371 GPa [13])
tific and technical significance owing to their ceramic-like and high hardness (about 16 GPa Vickers hardness [14])
mechanical behavior coupled with good electrical and ther- TiB whiskers as reinforcements in various titanium alloys.
mal conductivity. TiB and Ti3B4, are orthorhombic [2–4], The absence of intermediate phases between Ti and TiB
while TiB2 is hexagonal [3] in crystal structure. There is a as well as the possibility to grow TiB as high aspect ratio
large body of work on TiB2, but the interesting attributes whiskers in the Ti matrix (substantial increases in composite
of TiB compound are only beginning to be noticed. In tita- modulus and strength can be achieved this way) by simple
nium matrix containing small amounts of B, TiB forms heat treatments, are some of the unique advantages [12].
Further, the feasibility of making dense Ti–TiB functionally
*
Corresponding author. graded materials, with easily controllable gradation profiles
E-mail addresses: krutibas@gmail.com (K.B. Panda), ravi@mines. achieved by in situ growth of TiB in titanium, has also been
utah.edu (K.S. Ravi Chandran).
1
Address: Technology Transfer Office, University of Utah, Apartment
demonstrated [14]. While these works highlight the merits
#5, 120 North, G Street Salt Lake City, Salt Lake City, UT 84103, United of TiB as reinforcement and as engineering ceramic, the
States. Tel.: +1 801 918 9264; fax: +1 801 581 4937. fundamental elastic constants of this compound, which

1359-6454/$30.00 Ó 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2005.12.003
1642 K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657

determine the polycrystalline elasticity and anisotropy, are magnetic, optical, dielectric and superconducting properties
not known. The polycrystalline elastic moduli of TiB are of materials. For instance, studies of electronic structure of
also a subject of controversy recently [12] due to difficulties transition metal compounds, TiC [18], TiO2 [19] and the
in direct measurement. The principal objective of this work pressure-dependent nonlinear elasticity and the associated
is to determine the nine independent elastic constants of TiB phase transitions in earth-mantle compounds such as
as well as its polycrystalline moduli by first principles com- MgSiO3 [20] and MgO [21], have recently shown the success
putational calculations. of this approach. The relevant solid state properties can be
calculated from the ground state crystal energies corre-
1.1. Uncertainties in TiB elastic properties sponding to the equilibrium electron densities of the cluster,
molecular unit or the unit cell. The fact that solid state
The peritectic reaction that governs the formation of properties can be determined by using just the crystal struc-
TiB from Ti-rich liquid and TiB2 makes it difficult to grow ture and the atomic positions data is quite appealing. It is
TiB single crystals required for the unambiguous experi- also a viable alternative when experimental methods are
mental determination of the nine anisotropic elastic con- problematic, such as in the present case of TiB or any other
stants of the orthorhombic TiB phase. Atri et al. [13] such compound. Further, valuable insights can be gained
took the course of estimating the TiB polycrystalline elastic from electronic charge densities at the ground state, for
modulus values from elastic moduli data of Ti–TiB com- instance, the directionality in chemical bonding between
posites containing high volume fractions of TiB whiskers. atoms and its effect on elastic anisotropy and bulk proper-
The calculations yielded elastic moduli ranging from 325 ties. In this study, the DFT implementation in WIEN2K
to 435 GPa. Other works [15–17] either made estimates [22], a code that uses full-potential linear augmented plane-
by a similar procedure or arbitrarily assumed the polycrys- wave method (FPLAPW), was used to determine the equi-
talline modulus of TiB leading to hypothetical values that librium electronic structure and the anisotropic elastic
ranged from 482 to 550 GPa. While these values may seem constants of TiB. The calculations were performed with
to put TiB on par with SiC and Al2O3, the well known the generalized gradient approximation (GGA) to treat
engineering ceramics, the large variability in TiB modulus the exchange and correlation effects in DFT.
numbers, however, is a problem that needs to be resolved
before TiB may be accepted in applications, either as rein- 2. Methodology
forcement or as monolithic. Since experimental measure-
ments must await successful growth of TiB single 2.1. Crystal structure of TiB
crystals, it may be worthwhile to use the first principles
computational methods to determine the nine independent Decker and Kasper [2] were the first to establish that TiB
elastic constants and the polycrystalline elastic moduli of crystallizes in FeB structure, which is a primitive ortho-
TiB. Further, the presence of the continuous B chain along rhombic (Pnma) crystal belonging to the space group 62
the Y-direction of TiB suggests that TiB may be signifi- with mmm symmetry. The lattice parameters determined
cantly elastically anisotropic and the electronic structure by X-ray powder diffraction were a0 = 6.12 ± 0.01 Å,
details derived from computational calculations can pro- b0 = 3.06 ± 0.01 Å and c0 = 4.56 ± 0.01 Å, respectively.
vide valuable information in this respect. Fig. 1(a)–(d) illustrates various aspects of the crystal struc-
The present study first demonstrates the procedure ture of TiB. The fundamental building block of TiB is the
employed to determine the nine independent elastic con- trigonal prism with the B atom at the center and the Ti
stants of the orthorhombic TiB from first principles calcula- atoms in corners. The orthorhombic unit cell is formed by
tions on the basis of computational implementation of the the transverse stacking of the trigonal prisms in columnar
density functional theory (DFT). The calculated values arrays connected only at their edges (Fig. 1(a)) [2,3]. This
are then assessed in the light of experimental estimates made stacking leads to a zig-zag chain of B atoms along the
in our earlier work. The study also explores the electronic [0 1 0] direction with the atomic structure at the base as
nature of bonding between the atoms by investigating the shown in Fig. 1(b). This particular arrangement leads to
density of states in valence bands and the three-dimensional the TiB a needle-like whisker shape that develops due to
character of electron charge density distributions in differ- the anisotropic nature of B diffusion [9]. The orthorhombic
ent crystallographic directions, in order to understand the unit cell containing the 4Ti and 4B atoms as illustrated in
origins of high hardness and high elastic modulus as well Fig. 1(c). There is one B and one Ti in nonequivalent atomic
as the degree of elastic anisotropy of TiB. positions of Ti (0.177, 0.25, 0.123) and B (0.029, 0.25,
0.603). Fig. 1(d) illustrates three unit cells stacked in the
1.2. Density functional theory based computational Y-direction to illustrate the B–B chain.
determination of properties
2.2. Calculation of elastic constants
Computational calculation of ground state properties of
assemblies of atoms using first principles methods based on Crystals deform almost in a linear elastic manner at
DFT is a powerful tool for the determination of structural, small strains and the strain energy density increment of
K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657 1643

Fig. 1. Crystal structure of TiB: (a) illustration of the stacking of the trigonal prisms with zig-zag B chain [3], (b) illustrates the close packing of the prism,
(c) shows the unit cell of TiB containing 4Ti and 4B atoms and (d) illustrates the zig-zag B chain. In (c,d) the large atoms are Ti and the small atoms are B.

a homogeneous and elastically deformed crystal is given 2M 2


cii ¼ for the normal strains ði ¼ 1–3Þ and
by V0
ð5Þ
dE ¼ rij ddij ð1Þ M2
cii ¼ for the shear strains ði ¼ 4–6Þ
2V 0
where rij are the elements of stress tensor and dij are the ele-
ments of strain tensor. In a more generalized form, the sin- Eq. (5) implies that other coefficients, M0, M1 and M3 are
gle crystal elastic constants can be expressed as zero and this will be the case for pure linear elastic strain.
However, these coefficients may not be exactly zero in ac-
o2 E
¼ cijkl ð2Þ tual calculations, but can be made close to zero with extre-
odij odkl mely small strains. Therefore, small elastic strains, di,
According to Wallace [23], the internal energy of a crystal i = 1–6, can be applied to the undeformed unit cell lattice
under a general strain dij can be expressed by expanding the with the equilibrium atomic configuration and elastic con-
internal energy E(V) of the deformed crystal with respect to stants can be determined from the resulting change in
the strain tensor, in terms of the Taylor’s series, as energy. The nine independent elastic constants for the
X orthorhombic crystals are c11, c22, c33, c12, c23, c13, c44,
V0X
EðV Þ ¼ EðV 0 ;0Þ þ V 0 rij dij þ cijkl dij dkl þ  ð3Þ c55 and c66 are determined by selectively imposing strains
i;j
2! i;j;k;l either individually or in combination along specific crystal-
where V0 is the volume of the unstrained crystal and E(V0,0) lographic directions. Table 1 illustrates the deformations
is the corresponding ground state energy. In the above employed in determining c11, c22, c33, c44, c55 and c66 di-
expression the strain tensors subscripts (ij, kl) can be further rectly from deformations and c12, c13, c23 from a combina-
expressed in Voigt notation scheme (11 = 1, 22 = 2, 33 = 3, tion of deformations and other elastic constants.
23 = 4, 31 = 5 and 12 = 6). With this, Eq. (3) is of the form
2.3. Unit cell distortions, changes in symmetry and internal
EðV Þ ¼ M 0 þ M 1 di þ M 2 d2i þ    ð4Þ
atomic displacements
with M0 = E(V0,0), M1 = V0ri and M2 = (V0cii/2.) There-
fore, the second-order coefficient of a polynomial fit to Considering the R matrix containing the axis vectors for
strain energy density with respect to the deforming strain the undeformed Bravais lattice, any distortion matrix D
component should give the elastic constant for that strain: will move every Bravais lattice points of the undistorted
1644 K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657

Table 1
Distortion matrix and elastic stiffness for orthorhombic system (TiB)
Distortion matrix Crystal structure Space group (symmetry) No. of Relationship between elastic constant
after deformation after deformation symmetry and the coefficient of fit
operations
0 1
1þd 0 0
@ 0 1 0A Orthorhombic 62 (Pnma) 8 c11 ¼ 2M 2
V
0 0 1
0 1
1 0 0
@0 1þd 0A Orthorhombic 62 (Pnma) 8 c22 ¼ 2M 2
V
0 0 1
0 1
1 0 0
@0 1 0 A Orthorhombic 62 (Pnma) 8 c33 ¼ 2M 2
V
0 0 1þd
0 1
1 0 0
@0 1 dA Monoclinic 14 (P2_1/a) 4 c44 ¼ M 2
2V
0 d 1
0 1
1 0 d
@0 1 0A Monoclinic 11 (P2_1/m) 4 c55 ¼ M 2
2V
d 0 1
0 1
1 d 0
@d 1 0A Monoclinic 14 (P2_1/a) 4 c66 ¼ M 2
2V
0 0 1
0 1
1 þ 2d 0 0
@ 0 1d 0 A Orthorhombic 62 (Pnma) 8 4c11  4c12  4c13 þ c22 þ 2c23 þ c33 ¼ 2M 2
V
0 0 1d
0 1
1d 0 0
@ 0 1 þ 2d 0 A Orthorhombic 62 (Pnma) 8 c11  4c12 þ 2c13 þ 4c22  4c23 þ c33 ¼ 2M 2
V
0 0 1d
0 1
1d 0 0
@ 0 1d 0 A Orthorhombic 62 (Pnma) 8 c11 þ 2c12  4c13 þ c22  4c23 þ 4c33 ¼ 2M 2
V
0 0 1 þ 2d

lattice to a new position in the distorted R 0 matrix, which is c33 distortion of the TiB2 hexagonal unit cell. The volume
given by change is illustrated by the gray region (Fig. 2(b)). The
ground state crystal energy E(V0, d3) in the deformed con-
R0 ¼ RD ð6Þ
dition is calculated with the specification of new lattice
Here D is the distortion (for pure shear, rotation is ignored) parameters and internal atomic coordinates.
matrix containing the deformation strain. Table 1 describes The pure shear distortions involved in the determination
the distortion matrices employed, the crystal structure and of c44, c55 and c66 will change the symmetry of the ortho-
space group after deformation and the relationship be- rhombic crystal to new space groups in the monoclinic sys-
tween the elastic constant and the second-order coefficient tem (Table 1) depending on the type of deformation.
of the polynomial. Fig. 3(a) illustrates schematically the c55 shear deforma-
The six distortions involving c11, c22, c33, c12, c23, and c13 tion. Fig. 3(b) shows the three-dimensional view of the
do not change the orthorhombic symmetry but are accompa- atoms in the unit cell. Fig. 3(c) shows the positions of B
nied by a volume change. This is not, however, the case for and Ti atoms in both the deformed and the undeformed
distortions involving c44, c55 and c66. They make the dis- crystal with the exaggeration of atomic displacements for
torted lattice belong to a monoclinic system. The pure shears, clarity. The gray circles represent the atoms in the unde-
in addition to changing symmetry (Table 1) also change the formed crystal. The atoms in the deformed crystal are
atomic positions and the deformed positions have to be shown in black. Both the lattice parameters ‘a’ and ‘c’
found out for the ground state energy calculations. and the included angle change as a result of the shear c55.
The unit cell distortions involved in the determinations The shear strain d5 is
of c11, c22 and c33 preserve the symmetry of the unit cell
p  ou ow
but the internal atomic positions change according to the d5 ¼ 2exz ¼ b ¼ þ ð7Þ
distortion. As an example, Fig. 2(a) and (b) illustrate the 2 dz dx
K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657 1645

Fig. 2. c33 distortion in a orthorhombic TiB unit cell.

Fig. 3. Pure shear distortion d5 involved in c55 in the orthorhombic unit cell of TiB. (a) The schematic of shear distortion (b) the deformed state of the unit
cell as viewed in the Y-direction (c) schematic of Ti and B atom displacements (not all atoms in the same plane) and (d) illustration of parameters involved
in the definition of the shear strain d5.

The displacements of internal Ti and B atoms after shear, nal atom displacements required to impose the shear can be
into new deformed positions, were determined from the determined. It may be noted that in Eqs. (8) and (9), the
undeformed fractional coordinates, fa and fc, of an interior crystal was considered to deform symmetrically under pure
atom (B or Ti) in the X- and Z-directions, respectively. The shear, thus leading to (ou/dz) = (ow/dx) = exz with the ab-
internal atomic displacements in the X- and Z-directions, sence of rotations. Similarly, in case of other shear defor-
ou* and ox* due to the shear are then given by mations, c44, and c66, the symmetry of the lattice changes
  and the approach similar to that used for c55 was followed
 ou  1
ou ¼ c ¼ exz c ¼ d5 fc c ð8Þ to determine atomic displacements due to deformation.
dz 2
 
ow  1
ow ¼ a ¼ exz a ¼ d5 fa a ð9Þ 2.4. Computational details
dx 2
Thus, from lattice parameters, fractional coordinates of The ground state crystal energies of the undeformed
nonequivalent atoms and the deformation strain, the inter- and the deformed states of TiB unit cell under different
1646 K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657

distortions were determined using the computational


implementation of DFT, WIEN2K [22]. This implementa- -7029.8256
tion uses the FLAPW method that allows wave function
representation in the periodic crystal in the reciprocal

Energy (Rydbergs)
-7029.8258
space with a choice of relativistic or nonrelativistic calcu-
lations using the generalized gradient approximation
(GGA) for the electron density space to treat the exchange -7029.826
and correlation effects within DFT. This approach is com-
monly known as one of the most accurate schemes for
electronic structure calculations in solids. -7029.8262

In the DFT calculations, it is important to ensure that 50µRy


the ground state energy is converged within a specified tol- -7029.8264
erance in terms of self-consistent iterations. This means
that the electron density in the electron space around the 0 1000 2000 3000 4000 5000 6000 7000
nuclei has reached the equilibrium value self-consistently Number of k-points
with respect to all the interatomic forces. This can be
achieved only by a good numerical sampling of electron Fig. 4. K-mesh optimization for TiB.
densities in the Brillouin zone. This is largely dependent
on the number of sampling points for numerical integra- -7029.72
tions of electron densities in the k-space in the irreducible
Brillouin zone. A test was first made using experimental -7029.74
TiB unit cell parameters to determine the number of k-
Energy (Rydbergs)

points required. Fig. 4 illustrates the variation of ground -7029.76


state energies of equilibrium TiB unit cell as a function of
k-points. Above 5000 k-points and up to 6000 k-points, -7029.78
the changes in the ground state energies with the number
of k-points were about 30 lRy and therefore, 5000 points -7029.8
were used in the present set of calculations, so that exces-
sive computational time was not incurred. -7029.82
In the WIEN2K implementation of the FLAPW
-7029.84
method, the self-consistent ground state energy calcula-
tions are performed at two levels. Within core regions of
-7029.86
atoms, that is inside atomic spheres of radius, Rmt, a linear 5 6 7 8 9 10
combination of radial functions times spherical harmonics R *K
mt max
is used to determine the core electronic contributions to the
energy whereas in the interstitial regions, Fourier series Fig. 5. Rmt K max optimization for TiB.
representation of plane waves is used to determine the con-
tribution of valence electrons.2 The radius of the boundary electron densities. This accuracy was considered sufficient
between the two regions is the radius, Rmt and for the pres- for the present calculations. Relaxations of internal atoms
ent calculations, Rmt was taken as 1.6 Bohr. It is important in the deformed state were performed using the ‘‘PORT’’
to check the convergence of the crystal energy as a function module in WIEN2K until the interatomic forces between
the term RmtKmax, where Kmax is the magnitude of the larg- the atoms were less than 2 mRy/a.u.
est k-vector in the basis set. With a 5000 point k-mesh,
RmtKmax was varied within the range of 6.0–9.5 with a step 3. Results and discussions
size of 0.5. Fig. 5 illustrates the variation of crystal energy
as a function of RmtKmax. It can be seen that a value of 3.1. Structure optimization
RmtKmax = 7 or greater is a good approximation for the
basis set in the calculations without being computationally The structure parameters for TiB were first optimized in
very time consuming. All calculations of crystal energies WIEN2K to ensure that the calculations were based on
were made using the same set of parameters: Rmt = equilibrium lattice constants. The additional purpose is to
1.6 Bohr, RmtKmax = 7 with 5000 k-points. With these check whether or not the crystal energy minimum is
choices, the total energy was converged to within 10 lRy achieved for a given set of calculation parameters such as
per atom within 16–24 iterations for self-consistency of Rmt, k-points, as these are the only numerical parameters
that affect the accuracy of calculations. The experimental
2
Complete details of this method is documented in the WIEN2K
lattice parameters (Table 2) were taken as the basis for
reference manual, published by P. Blaha, K. Schwarz, J. Luitz, WIEN2K, doing the optimization using the energy minimization crite-
Vienna University of Technology, Vienna, Austria. ria following the orthorhombic symmetry. The volume was
K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657 1647

Table 2 3.2. TiB single crystal elastic constants


The experimental and equilibrium structure parameters (GGA) of TiB
a (Å) b (Å) c (Å) Volume (Å3) b/a c/a The nine independent elastic constants for TiB were
Experimental 6.12 3.06 4.56 85.396 0.5 0.745 determined by imposing nine different deformations given
Present calculation 6.111 3.051 4.559 85.001 0.499 0.746 in Table 1 on the equilibrium orthorhombic unit cell and
determining the dependence of the resulting change in strain
optimized first. In WIEN2K, this is performed by applying energy on the deformation. Up to seven deformations, 2%,
hydrostatic pressure by changing the lattice parameters 1%, 0.5%, 0%, 0.5%, 1%, and 2%, were applied to deter-
while keeping the ratios, c/a and b/a, as constants. The vol- mine each elastic constant. The ground state energies of TiB
ume variations applied were 10%, 5%, 0%, +5%, and unit cell were determined for all these deformation strains.
+10% of the experimental equilibrium volume. With the Fig. 7(a)–(i) illustrate the total energy as a function of strain
optimized volume, the ratios, c/a and b/a, were varied to for the nine distortions. The lines in the figures represent the
find the values corresponding to the crystal energy mini- second-order polynomial fit (Eq. (4)) to the data from which
mum. Fig. 6 illustrates the crystal energy variations with the values of M2 were determined. The changes in energy
respect to the structure parameters. The crystal parameters with deformation before internal atomic relaxation are
corresponding to the minimum energy are listed in Table 2. shown as dotted lines. The elastic constants were then
There is very good agreement between the values obtained calculated using equations in Table 1. Among the nine inde-
using WIEN2K and the lattice parameters optimized by pendent elastic constants, determination of c44, c55 and c66
others. For example, Mizuno et al. [33] determined the lat- was the most time consuming probably because they involve
tice parameters, a, b and c of TiB crystal as 6.112, 3.054, monoclinic unit cell with a reduced symmetry (Table 1)
4.56 Å, respectively. These compare very well with the val- resulting from the deformation. The calculated elastic con-
ues of present calculation presented in Table 2. stants are presented in Table 3.

0.05
a 3
Expt. V = 85.396032 A
0
3
Calc. V = 85.0014252 A
0
0.04

0.03
δE (Ry)

0.02

0.01

75 80 85 90 95
Volume (A /Cell)
3

0.0025 0.0035
b Expt. c/a = 0.745098 c Expt. b/a = 0.5
Calc. c/a = 0.746109 0.003 Calc. b/a = 0.4993684
0.002
0.0025
0.0015
0.002
δE (Ry)

δE (Ry)

0.001 0.0015

0.001
0.0005
0.0005
0
0

-0.0005 -0.0005
0.73 0.735 0.74 0.745 0.75 0.755 0.76 0.765 0.485 0.49 0.495 0.5 0.505 0.51 0.515
c/a b/a
Fig. 6. The structure optimization plots using GGA calculations. (a) Volume optimization. (b) c/a optimization. (c) b/a optimization.
1648 K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657

0.0035 0.005
a Unrelaxed b Unrelaxed
0.003 Relaxed Relaxed
0.004
0.0025
0.003
0.002

δE(Ry)
δE(Ry)

0.0015 0.002

0.001
0.001
0.0005
0
0
C /2 C /2
11 22
-0.0005 -0.001
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 -0.03 -0.02 -0.01 0 0.01 0.02 0.03
δ δ

0.004
c 0.012
Unrelaxed d Unrelaxed
Relaxed 0.01 Relaxed
0.003

0.008
0.002
δE(Ry)
δE(Ry)

0.006

0.001 0.004

0.002
0
0
C /2 2C
33 44
-0.001 -0.002
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 -0.03 -0.02 -0.01 0 0.01 0.02 0.03
δ δ

0.015
e
Unrelaxed
Relaxed
0.01
δE(Ry)

0.005

2C
55
-0.005
-0.03 -0.02 -0.01 0 0.01 0.02 0.03
δ

Fig. 7. (a)–(i) Strain energy of deformation as a function of deformation parameter for nine different deformations of TiB crystal.

3.3. Effect of relaxation of atoms correspond to the equilibrium atomic state of the deformed
unit cell, as the B–B, Ti–B and Ti–Ti bond distances were
The elastic constants in the first row of Table 3 are not allowed to equilibrate. Since it is natural for the atoms
termed unrelaxed because the atomic coordinates of the to adjust their positions to maintain optimum bond dis-
deformed unit cell strictly followed the linear elastic defor- tances, the internal B and Ti atoms must be allowed to
mations imposed. However, this does not necessarily relax to find a lower energy minimum of the crystal in
K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657 1649

0.01 0.02
f Unrelaxed g Unrelaxed
Relaxed Relaxed
0.008
0.015

0.006
0.01
δE(Ry)

δE(Ry)
0.004
0.005
0.002

0
0
2C (4C -4C -4C +C +2C +C )/2
66 11 12 13 22 23 33
-0.002 -0.005
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 -0.03 -0.02 -0.01 0 0.01 0.02 0.03
δ δ

0.025 0.025
h Unrelaxed i Unrelaxed
Relaxed Relaxed
0.02 0.02

0.015 δE(Ry) 0.015


δE(Ry)

0.01 0.01

0.005 0.005

0 0
(C -4C +2C +4C -4C +C )/2 (C +2C -4C +C -4C +4C )/2
11 12 13 22 23 33 11 12 13 22 23 33
-0.005 -0.005
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 -0.03 -0.02 -0.01 0 0.01 0.02 0.03
δ δ
Fig. 7 (continued)

Table 3
The calculated single crystal elastic constants and the stiffness matrix for
orthorhombic TiB before and after relaxation
δE

Elastic constant c11 c22 c33 c12 c13 c23 c44 c55 c66
Unrelaxed (GPa) 410 516 457 84 79 54 334 367 291 unrelaxed
Relaxed (GPa) 411 524 410 91 107 61 189 186 193 Relaxation Space
δE

the relaxation space as schematically illustrated in Fig. 8.


This is particularly important for shear deformations.
The true relaxation of atoms involves adjustments of coor-
dinates in all directions while keeping the deformed unit relaxed
cell shape fixed. Additionally, the symmetry of the unit cell
should not change from that of the unrelaxed and ε
deformed configuration as this may push the crystal to Fig. 8. A schematic illustrating the method of relaxation. The dotted
some other minimum energy configuration that would have curve is the plot of energy vs. deformation after relaxation. The lowest
no relation to the original deformed state. It may be value of energy is obtained from the minima in the relaxation space.
recalled that except for the three shear deformations c44,
c55 and c66, the orthorhombic symmetry is preserved in
the rest of the deformations. Although, in principle, sym- internal atom positions in the orthorhombic TiB unit cell,
metry preserving deformations have limited scope to relax, it is evident that because the Y-positions of both Ti and
this is only true for systems where internal atoms are sym- B atoms, are fixed by symmetry requirement, the atoms
metry constrained to occupy fixed positions. From the are free to move only in the XZ-plane. This means that
1650 K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657

even though there is no reduction in symmetry in deforma- Table 4 lists the forces on atoms before and after relax-
tions involving c11, c22, c33, and c12, c13, c23, there may be ation for c33 and c55 deformations. Table 5 lists the frac-
relaxation from the free movement of the Ti and B atoms tional coordinates of atoms before and after relaxations
in the XZ-plane, in these deformations. On the other hand, for these two deformations. It can be seen that for c33,
in deformations involving the shear deformations, c44, c55 the largest force on B atom was in the Z-direction
and c66, the symmetry is reduced to monoclinic and the (16.113 mRy/a.u.), which is completely relaxed when the
Y-positions of the internal Ti and B atoms are not fixed fractional Z-position of this B atom was reduced from
by symmetry and hence they are free to move in the plane 0.603 to 0.595 (Table 5). This means that the B–B bond dis-
normal to the shear axis. tance is slightly reduced in the relaxed configuration. It
The PORT module in WIEN2K was used to find the may be noted that the Fy in Table 4 for all atoms are zero.
global minimum of the total energy by minimizing the In c55, the crystal symmetry is reduced to monoclinic and
interatomic forces with the help of the first-order deriva- there are four nonequivalent atoms in the unit cell (Table
tives of energy with respect to the atomic positions. The 5). Since the shear is about Y-direction, the B–B distance
relaxations of the atoms were performed until the inter- in Y-direction does not change, thus the FY values has to
atomic forces were less than 2 mRy/a.u. The changes in be zero on all atoms. More importantly, the X- and Z-coor-
crystal energy with deformation and after relaxation are dinates of Ti and B atoms in the deformed monoclinic cell
shown as solid lines in Fig. 7(a)–(i). The elastic constants actually seem to approach the corresponding coordinates
determined from the polynomial fits of relaxed energies of the undeformed orthorhombic cell. This means that dur-
are given in the second row of Table 3. It can be seen that ing relaxation, the crystal tries to restore itself to the origi-
there are substantial changes in the values of the elastic nal undeformed orthorhombic lattice, at least with respect
constants. The relaxation causes almost no change in c11, to the internal atoms. This should reduce the crystal
a slight increase in c22, and a significant reduction in c33. ground state energy significantly and thus, a large decrease
The decrease in elastic constants due to relaxations in pure in c55 is seen after relaxation. Similar changes in atomic
shears, c44, c55 and c66 are substantially higher in compar- positions occur in other shear constants to the extent
ison to the other elastic constants. This is understandable reflected in the reductions of elastic constants.
considering the fact that the reduction in symmetry elimi-
nated the Y-position constraint leading to a higher degree 3.4. Elastic constants of TiB polycrystalline aggregate
of freedom to relax in the shear deformations. The extent
of relaxation in c55 is greater than that in c44 and c66. This The theoretical polycrystalline elastic modulus for TiB
is because of the fact that shear deformation c55 involves can be determined from the set of nine independent elastic
torque about the B–B bond and it is likely that the B–B stiffness constants. There are two approximation methods
bond strength is higher than Ti–B and Ti–Ti bonds. Here, to calculate the polycrystalline modulus, namely, the
a small change in B atom position can lead to a large Voigt method [24] and the Reuss method [25]. They pro-
change in the elastic constant. vide the upper (Voigt) and lower (Reuss) bounds to the

Table 4
Forces on atoms before and after relaxation for c33 and c55 deformations
Paramater Atom Forces (unrelaxed) (mRy/a.u) Forces (relaxed) (mRy/a.u)
FX FY FZ FX FY FZ
c33 Ti 2.832 0 2.019 0.473 0 0.11
B 1.127 0 16.113 0.534 0 0.021
c55 Ti1 6.368 0 4.387 0.023 0 0.131
B1 11.873 0 8.725 0.342 0 0.727
Ti2 5.124 0 1.432 0.349 0 0.064
B2 0.463 0 1.934 0.873 0 0.243

Table 5
Atomic positions before and after relaxation with +2% deformations for c33 and c55
Deformation Unrelaxed atom positions (X/a, Y/b, Z/c) Relaxed atom positions (X/a, Y/b, Z/c) Energy (Ry.) (unrelaxed) Energy (Ry.) (relaxed)
c33 Ti (0.177, 0.25, 0.123) Ti (0.1771, 0.25, 0.1242) 7029.822459 7029.824432
B (0.029, 0.25, 0.603) B1 (0.029, 0.25, 0.595)
c55 Ti1 (0.1787, 0.25, 0.1277) Ti1 (0.1771, 0.25, 0.1223) 7029.807341 7029.808903
Ti2 (0.6826, 0.25, 0.3722) Ti2 (0.6774, 0.25, 0.3778)
B1 (0.0230, 0.25, 0.6317) B1 (0.0313, 0.25, 0.5931)
B1 (0.5275, 0.25, 0.8843) B1 (0.5289, 0.25, 0.8923)
K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657 1651

Table 6
Isotropic values of bulk modulus, shear modulus and elastic modulus for single crystal TiB using Voigt, Reuss and Hill’s approximations
BV (GPa) BR (GPa) BH (GPa) GV (GPa) GR (GPa) GH (GPa) E (GPa) m
Unrelaxed 202 201 201.5 276 255 265.5 553 0.04
Relaxed 207 206 206.5 186 183 184.5 427 0.15

polycrystalline elastic modulus. For orthorhombic crys- tions. Clearly, higher values of this range, 550 GPa is
tals, the shear modulus (G) and the bulk modulus (K) unrealistic since the polycrystalline E of TiB2 itself is about
according to Voigt and Reuss approximations are given 570 GPa [27]. It is likely that the polycrystalline E of TiB is
by3 significantly lower than this, since the number of B–B
1 1 bonds and their continuity is lower in TiB relative to TiB2.
GV ¼ ðc11 þ c22 þ c33 Þ  ðc12 þ c13 þ c23 Þ
15 15
1 3.5. Comparison with estimates from experimental data
þ ðc44 þ c55 þ c66 Þ ð10Þ
15
1 4 4 Some indirect experimental evidence is available to
¼ ðs11 þ s22 þ s33 Þ  ðs12 þ s13 þ s23 Þ assess the accuracy of first principles calculations of TiB
GR 15 15
3 polycrystalline elastic modulus values determined here.
þ ðs44 þ s55 þ s66 Þ ð11Þ Atri et al. [13] measured elastic moduli of titanium compos-
15
ites containing high volume fractions (ranging from 0.3 to
1 2
K V ¼ ðc11 þ c22 þ c33 Þ þ ðc12 þ c23 þ c13 Þ ð12Þ 0.83) of TiB whiskers created in situ. Measurements of res-
9 9 onant frequencies were made on dense rectangular slabs of
1
¼ ðs11 þ s22 þ s33 Þ þ 2ðs12 þ s23 þ s13 Þ ð13Þ these composites using the Grindosonic unit under the
KR pulse-induced vibration mode. The polycrystalline E values
in which the subscripts V and R indicate the Voigt and Re- of these composites, calculated from the resonant frequen-
uss averages. The arithmetic average of the Voigt and the cies (Table 7), are plotted as a function of TiB volume frac-
Reuss bounds are called the Voigt–Reuss–Hill (VRH) aver- tion in Fig. 9(a). The two lines represent calculations on the
age and it is used as the best estimate of the theoretical basis of Halpin–Tsai equation [28] for TiB whisker aspect
polycrystalline elastic modulus [26]. The VRH averages ratios of 1 and 20 and based on ETi = 110 GPa and a guess
are given by of ETiB = 371 GPa. The choice of 371 GPa as the polycrys-
1 1 talline E value of TiB was to provide the best fit to the com-
G ¼ ðGR þ GV Þ K ¼ ðK R þ K V Þ ð14Þ posite modulus data as a function of TiB volume fraction.
2 2
The isotropic bulk modulus and shear modulus values for Table 7
polycrystalline TiB are calculated using the above equa- Polycrystalline Ti–TiB composite Young’s moduli, shear moduli and
Poisson’s ratio [data taken from Ref. [13]]
tions and the results are given in Table 6. The polycrystal-
line elastic modulus and the Poisson’s ratio can be Material Elastic modulus Shear modulus Poisson’s
(E) (GPa) (G) (GPa) ratio (t)
computed from these values using the relationships:
Ti 117a 44a 0.32a
9KG 3K  2G 117b 43b 0.35b
E¼ t¼ ð15Þ
3K þ G 2ð3K þ GÞ 119c
Ti–30TiB 160a 62a 0.29a
Table 6 also lists the calculated values of polycrystalline 160b 62b 0.29b
elastic modulus and Poisson’s ratio of TiB based on the re- 169c
laxed values of the elastic constants. The polycrystalline
Ti–54TiB 210a 84a 0.25a
elastic modulus, shear modulus and Poisson’s ratio for 210b
TiB are 427 GPa, 185 GPa and 0.15, respectively, calcu- 213c
lated here using Eqs. (10)–(15). These values provide a the-
Ti–69TiB 276a 113a 0.22a
oretical reference to evaluate the reported elastic modulus 276b
values of polycrystalline TiB aggregate determined using 255c
different experimental extrapolation methods [13,15–17]. Ti–83TiB–7TiB2 367a 155a 0.17a
The reported elastic modulus values of isotropic homoge- 366b 155b 0.18b
neous polycrystalline TiB ranges from 371 to 550 GPa 356c
and were obtained from the moduli of several Ti–TiB com- TiB–16TiB2 425a 185a 0.15a
posites based on different estimation methods or assump- 426b
408c
3 a
The lower (Reuss) bounds of the polycrystalline elastic modulus were Tests conducted at JW Lemmens, St. Louis, MO.
b
calculated from the compliance constants, sij, determined by the inversion Tests conducted at LoTec, Salt Lake City, UT.
c
of cij values determined here. Tensile test data.
1652 K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657

500 200
a b GTiB = 169 GPa
ETiB = 371 GPa

400

Shear Modulus, G (GPa)


Elastic Modulus, E (GPa)
150

300

100

200

Dynamic 50
100 Tensile
Dynamic-1
Halpin-Tsai (l /d = 1) Dynamic-2
f f Halpin-Tsai (l f/df = 1)
Halpin-Tsai (l /d = 20)
f f Halpin-Tsai (l /d = 20)
f f
0 0
0 0.2 0.4 0.6 0. 8 1 0 0.2 0.4 0.6 0.8 1
Volume Fraction of TiB Volume Fraction of TiB

Fig. 9. Experimentally extrapolated values of polycrystalline. (a) Elastic modulus, and (b) shear modulus for TiB, determined from the experimental data
of Ti–TiB composites with various volume fractions of TiB phase. The data are from Ref. [13].

Therefore, 371 GPa is considered as the experimentally degrees of anisotropy in atomic bonding in different planes.
extrapolated value of polycrystalline modulus of TiB. It The shear anisotropic factors are given by
is quite possible that this value provides a lower bound esti- 4c44
mate, because for the highest TiB volume fraction compos- A1 ¼ for the f1 0 0g plane ð16Þ
c11 þ c33  2c13
ite (TiB–16TiB2), the dynamic elastic modulus values are
4c55
about 425 GPa. The first principles estimate (Table 6) A2 ¼ for the f0 1 0g plane ð17Þ
based on relaxed figurations of atoms yields 427 GPa, c22 þ c33  2c23
which is although about 15% higher than the estimated 4c66
A3 ¼ for the f0 0 1g plane ð18Þ
experimental value, it is quite close to the modulus value c11 þ c22  2c12
of TiB–16TiB2 composites. The calculated values of A1, A2 and A3 for TiB are given in
Fig. 9(b) illustrates the plot of the shear moduli of high Table 8. A value of unity means that the crystal exhibits
volume fraction Ti–TiB composites, estimated from the isotropic properties while values other than unity represent
torsional resonant frequencies measured using the Grindo- varying degrees of anisotropy. From Table 8, it can be seen
sonic unit. The data points are experimental values and the that TiB exhibits low anisotropy. Another way of measur-
solid lines are the best fit to the composite shear moduli, ing the elastic anisotropy is given by the percentage of
assuming GTi = 41 GPa and GTiB = 169 GPa for two TiB anisotropy in the compression and shear [29,30]. They are
whisker aspect ratios. The first principles estimate of GTiB defined as
is 185 GPa (Table 6) determined here is about 9% higher
KV  KR
than the experimentally extrapolated GTiB value. The Pois- Acomp ¼  100 ð19Þ
son’s ratio obtained by experimental extrapolation is 0.17 KV þ KR
which is somewhat higher than 0.15 calculated here (Table GV  GR
Ashear ¼  100 ð20Þ
6). However, many of the extrapolations used here are GV þ GR
based on certain assumptions which may not be entirely For crystals, these values can range from zero (isotropic) to
accurate for TiB polycrystals. Therefore, direct experimen- 100% representing the maximum anisotropy. The percent-
tal measurements on polycrystalline TiB may be needed to age anisotropy values were computed for TiB and are
verify the Poisson’s ratio. shown in Table 8. It can be seen that the anisotropy both
It is to be noted that our recent WIEN2K-DFT calcula- in shear and compression is also small.
tions [27] of elastic constants of TiB2 were in very good An illustrative way of describing the elastic anisotropy is
agreement with the experimental data on TiB2 single crys- a three-dimensional surface representation showing the
tals. Drawing on these conclusions, we believe the calcu- variation of elastic modulus with crystallographic direc-
lated values of ETiB = 427 GPa and GTiB = 185 GPa are tion. This directional dependence of the Young’s modulus
quite accurate. for orthorhombic crystals is given by [31]
Table 8
3.6. Anisotropy in TiB
Various measures of anisotropy in TiB as determined by calculations of
the present study
Many low symmetry crystals exhibit a high degree of
Method A1 A2 A3 Acomp (%) Ashear (%)
elastic anisotropy. The shear anisotropic factors on differ-
GGA 1.2454 0.9162 1.0252 0.3233 0.6836
ent crystallographic planes provide a measure of the
K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657 1653

Fig. 10. (a) Directional dependence of the elastic modulus in TiB. The plane projections of the directional dependence of the elastic modulus are shown in
(b) for comparison. The units are in GPa.

1 cant in-plane elastic anisotropy in XY- and YZ-plane is re-


¼ l41 s11 þ l42 s22 þ l43 s33 þ 2l21 l22 s12 þ 2l21 l23 s13
E vealed in Fig. 10(b). This is consistent with the atomic
þ 2l22 l23 s23 þ l21 l22 s66 þ l21 l23 s55 þ l22 l23 s44 ð21Þ arrangements in TiB. The B–B atom chain propagates in
the Y-direction [0 1 0] and considering B–B bond strength
where s11, s22, etc., are elastic compliance constants and l1, to be strongest, one would expect much less anisotropy in
l2 and l3 are the direction cosines. Fig. 10(a) and (b) illus- the plane perpendicular to the B chain (XZ-plane) while
trates the directional dependence of the TiB elastic modu- a significant anisotropy should be seen in planes having
lus calculated using the elastic constants computed from the B–B chain (XY and YZ planes).
the present calculation. The degree of deviation in shape The directional dependence of the bulk modulus can be
from a sphere indicates the degree of anisotropy in the sys- calculated using the following relationship [31]:
tem since for an isotropic system one would see a spherical
1
shape. Fig. 10(a) shows an appreciable deviation from a ¼ ðs11 þ s12 þ s13 Þl21 þ ðs12 þ s22 þ s23 Þl22 þ ðs13 þ s23 þ s33 Þl23
spherical shape and hence one can conclude that TiB does K
exhibit some degree of anisotropy in elastic modulus. The ð22Þ
projections of the elastic constants on the XY, YZ, and Fig. 11(a) and (b) illustrate the directional dependence of
XZ planes are shown in Fig. 10(b). Although in-plane the bulk modulus calculated from Eq. (22). Similar to
anisotropy is almost nonexistent in the XZ-plane, signifi- the elastic modulus, the shape of the bulk modulus

Fig. 11. (a) Directional dependence of the bulk modulus in TiB. The plane projections of the directional dependence of the bulk modulus are shown in (b)
for comparison. The units are in GPa.
1654 K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657

representation deviates from the spherical shape although as well as Ti-3d levels below the EF, while largely due to
the extent of deviation is not as much as that of the elastic Ti-3d above EF. This is similar to the observation made
modulus. Fig. 11(b) shows the projections of the bulk mod- by the authors for the titanium diboride (TiB2) [27] and
ulus on different crystallographic planes. In the case of bulk by Mizuno et al. [33]. In addition to the difference in the
modulus, the in-plane anisotropy in bulk modulus in XY- nature of pseudo gap, the additional difference is the extent
plane is largely absent while that in XY- and XZ-plane it of hybridization between the Ti-3d and B-2p states, with
is significant. It is shown to be significant in plane contain- the hybridization being more pronounced in TiB2 than in
ing B–B chains (YZ-plane). The strong B–B bonds might TiB. Additionally, the charge transfer between Ti and B
be the reason behind the large exhibition of anisotropy in in TiB2 was found to be higher leading to significant
the YZ-plane. inter-sheet bonding in TiB2, relative to TiB. Another differ-
ence between TiB and TiB2 is the length of the B–B bond.
3.7. Analysis of chemical bonding The B–B bond distance in case of TiB2 (1.7493Å) is signif-
icantly smaller than that of TiB (1.8301 Å). The higher
Examination of the nature of chemical bonding, espe- degree ofTi-3d and B-2p hybridization as well as the stron-
cially the distribution of valence charges between atoms ger and shorter B–B in TiB2 seems to explain the higher
is necessary to explain the small overall elastic anisotropy melting point and higher stiffness of TiB2 relative to TiB.
and large in-plane anisotropies in some planes as well as In order to illustrate the three-dimensional electron den-
the large values of elastic constants in some directions. In sity topography, three planes in which the Ti–Ti, Ti–B and
chemical terms, the electro-negativity for Ti (1.54) and B the B–B bonds are coplanar were chosen to assess the
(2.04) suggests directional bonding with bulk of the valence topology of electron density in the interstitial space
electrons strongly localized around the B atoms. In order between the atoms. Figs. 13–15 illustrate the valence elec-
to explore the nature of bonding between the B–B, Ti–B, tron charge density maps in different crystallographic
and Ti–Ti atoms in the TiB compound, the density of states planes in terms of contour plots and 3D plots of charge
(DOS) for the valence bands and the electron charge den- densities. Fig. 13(a) illustrates the valence charge density
sity maps along different crystallographic planes were con- distribution in the (4 0 0) plane. The (4 0 0) plane consists
structed from the electron density data of the undeformed of two Ti and two B atoms. The topology of valence elec-
crystal. tron density distribution4 is also shown in Fig. 13(b) as a
Fig. 12 illustrates the total density of states and the pro- 3D plot to reveal the three-dimensional nature of electron
jected density of states for Ti-3d, hybridized Ti-4sp, B-2s charge distribution. It can be seen that B accumulates
and B-2p orbitals. The valley in DOS slightly to the right valence electrons, relative to Ti, as suggested by its elec-
of the Fermi level (EF) is the pseudo gap, which is not pro- tro-negativity. Further, there appears to be a significant
nounced here in TiB as much as it is in TiB2 [27,32]. In fact, accumulation of electrons across the Ti–Ti bond, but in
the continuity of total DOS, largely dominated by Ti-3d the direction of B–B bond, with the electron charges seem-
states exists across the pseudo gap, highlighting an impor- ingly belonging to B–B bond. This bond is actually between
tant difference in TiB electronic structure relative to that of one of the B atoms in the (4 0 0) plane and the other B atom
TiB2. In Fig. 12, it can be seen that strong hybridization that is not shown. Consistent with this, it is also seen that
between the Ti-3d and B-2p states occur near E = 1 eV. there is a relatively weak bonding between Ti–Ti and Ti–B
The density of states is largely contributed by both B-2p atoms in this plane as indicated by Fig. 13(a).
The Ti–Ti bond and the nature of valence charge density
in this region can be seen in more clarity in Fig. 14.
12 The ð6 7 1 3 7 7 2 0 2Þ crystallographic plane shown in
Total Fermi Level (Ef) Fig. 14(a) also contains two Ti and two B atoms.
10 Ti 4sp Fig. 14(b) shows the three-dimensional details of the elec-
Ti 3d
B 2s tron charge distribution. Similar to that found in (4 0 0)
8 B 2p plane, there does not seem to be any significant accumula-
DOS (/eV)

tion in the electron charge between Ti and Ti atoms. The


6 trench between the Ti atoms does not show electronic
charge density accumulation in the direction of the Ti–Ti
4 bond, thus indicating the lack of a strong bonding between
the Ti atoms. However, there is a small charge accumula-
2 tion in that region, but in the direction of the B–B bond
between the B atoms in this plane, although this B–B bond
0 (inter-chain bond) does not seem to be as strong as the
-12 -8 -4 0 4
Energy (eV)
4
Fig. 12. The total and the projected density of states in TiB determined in The valence charge density was obtained by subtracting the charge
the present study. density of the core states from the total charge densities distribution.
K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657 1655

Fig. 13. (a) Valence electron charge density distribution in the (4 0 0) plane in the TiB unit cell illustrating the Ti–B bond and (b) 3D plot showing the
topology of charge density in the (4 0 0) plane.

Fig. 14. (a) Valence electron charge density distribution in the ð6 7 1 3 7 7 2 0 2Þ plane in the TiB unit cell illustrating the Ti–Ti bond and (b) 3D plot
showing the topology of charge density in the ð6 7 1 3 7 7 2 0 2Þ plane.

Fig. 15. (a) Valence electron charge density distribution in the (1 2 9 0 3 6 7 3 9 7) plane in the TiB unit cell illustrating the B–B bond and (b) 3D plot
showing the topology of charge density in the (1 2 9 0 3 6 7 3 9 7) plane.
1656 K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657

other B–B bond (along the B–B chain), which is illustrated (3) The calculations yielded polycrystalline TiB elastic
in Fig. 15. modulus value of 427 GPa and shear modulus value
The nature of B–B bonding between the B atoms form- of 185 GPa which are reasonably close to that
ing the zig-zag chain in the Y-direction, as seen in the obtained in approximate experimental estimates
(1 2 9 0 3 6 7 3 9 7) plane, is illustrated in Fig. 15(a). This (ETiB = 371 GPa and GTiB = 169 GPa). However,
plane contains two B atoms having the shortest interatomic better experimental data is needed to clearly assess
distance (thus forming the local segment of the B–B chain) the validity of theoretical data obtained in this work.
and the two neighboring Ti atoms. The three-dimensional (4) TiB exhibits a small degree of anisotropy as calcu-
valence charge density distribution in this region is illus- lated by standard definitions of anisotropy. However,
trated in Fig. 15(b). A striking aspect is the observation significant in-plane elastic and bulk anisotropies
of valence charge density accumulation between the B–B exist, presumably caused by the large difference in
atoms in this plane and the highly directional nature of this bond strengths of B–B and Ti–B bonds.
bond. This suggests a strong covalent interaction between (5) Most of the electron density is accumulated between
the B and B atoms having the shortest bond distance along B–B bonds and in the bond direction. The largest
the B–B chain. Similar to that found in other planes, the accumulation is along the B–B bonds forming the
degree of charge accumulation is very small and the nature zig-zag chains and the next largest is along the
of bonding between Ti–B and Ti–Ti is much weaker that of inter-chain B–B bonds that cut across the Ti–Ti
B–B. Therefore, it seems much of the valence charge densi- bonds. All B–B bonds show a relatively higher accu-
ties in TiB are first accumulated along the shortest B–B mulation of valence charge density than either Ti–B
bond that is part of the B–B chain (Fig. 15(b)) and then or Ti–Ti bonds. This seems to explain the high mod-
the other short B–B that forms the inter-chain link ulus values of the TiB compound.
(Fig. 13(b)), with the Ti–Ti and Ti–B bonds seeing a very
little share of the valence charge distribution. This suggests
a quasi-continuous network of B–B bonds not only along Acknowledgments
the B–B zig-zag chain, but also between the B atoms
between the chains, although this link is not as strong as The authors thank Drs. P. Blaha, Vienna University of
that along the chain. This is somewhat analogous to the Technology, Austria, and Stefaan Cottenier, Institut voor
planar network of B–B bonds in TiB2, although, qualita- Kernen Stralingsfysica, Belgium for providing several clar-
tively the overall intensity of network is relatively smaller ifications during the calculations. Victor Bazterra of Center
in TiB than in TiB2, for, B in TiB2 has three nearest neigh- for High Performance Computing (CHPC), University of
bors unlike TiB. This perhaps provides a qualitative ratio- Utah, provided invaluable help with respect to WIEN2K
nale for the high stiffness of TiB, but not as high as that of code installation.
TiB2.
Among the single crystal elastic constants calculated, the References
magnitude of c22 is larger relative to c11 and c33. The short-
est B–B bonds, where an enhanced accumulation of valence [1] Baker H, editor. ASM handbook: alloy phase diagrams, vol.
electron density is seen are the ones that constitute the zig- 3. Materials Park (OH): ASM International; 1992. p. 2.
zag chain. These bonds are more aligned toward the Y- [2] Decker BF, Kasper R. Acta Crystallogr 1954;7:77.
[3] Fan Z, Guo ZX, Cantor B. Composite A 1996;28:131.
direction than to the X- and Z-directions of the TiB unit
[4] Fenish RG. Trans Metall Soc AIME 1966;236:804.
cell. Thus it is not surprising that the numerically largest [5] Lundstrom T. In: Matkovich VI S, editor. Boron and refractory
elastic constant is c22, since the bonds would be the stiffest borides. Berlin: Springer-Verlag; 1977. p. 351.
in this direction. [6] Saito T, Furuta T, Yamaguchi T. Development of low cost titanium
matrix composite. In: Froes FH, Storer J, editors. Recent advances in
titanium metal matrix composites. The Minerals, Metals and Mate-
4. Conclusions
rials Society; 1995. p. 33.
[7] Godfrey TMT, Goodwin PS, Ward-Close CM. Adv Eng Mater
(1) The nine independent elastic constants of orthorhom- 2000;2:85.
bic TiB single crystal compound have been calculated [8] Hyman ME, McCullough C, Levi CG, Meherabian R. Metall Mater
using DFT–FLAPW–GGA method. The calculated Trans A 1991;22:1647.
[9] Sahay SS, Ravichandran KS, Atri R. J Mater Res 1999;14:4214.
values in GPa are c11 = 411, c22 = 524, c33 = 410,
[10] Ravi Chandran KS, Panda KB. Adv Mater Process 2002(October):59.
c12 = 91, c13 = 107, c23 = 61, c44 = 189, c55 = 186 [11] Panda KB, Ravi Chandran KS. Metall Mater Trans A 2003;34:
and c66 = 193. 1371.
(2) Internal relaxation of atoms greatly influences the [12] Ravi Chandran KS, Panda KB, Sahay SS. JOM 2004(May):42.
single crystal elastic constants. The largest changes [13] Atri RR, Ravichandran KS, Jha SK. Mater Sci Eng A 1999;271:
150.
in elastic constants were seen in the symmetry reduc-
[14] Panda KB, Ravi Chandran KS. Metall Mater Trans A 2003;34: 1993.
ing shear deformations, c44, c55 and c66, where the B [15] Gorsse S, Miracle DB. Acta Mater 2003;51:2427.
atom tended to move close to the undeformed [16] Fan Z, Miodownik AP, Chandrasekaran L, Ward-Close M. J Mater
positions. Sci 1994;29:1127.
K.B. Panda, K.S. Ravi Chandran / Acta Materialia 54 (2006) 1641–1657 1657

[17] Tsang HT, Chao CG, Ma CY. Scripta Mater 1997;9:1359. [26] Hill R. Proc Phys Soc London Sect A 1952;65:349.
[18] Blaha P, Schwarz K, Redinger J. J Phys Met Phys F 1985;15:263. [27] Panda KB, Ravi Chandran KS. Comp Mater Sci 2006;35:134.
[19] Jhi S, Ihm J, Louie SC, Cohen ML. Nature 1999;399:132. [28] Mallick PK. Fiber-reinforced composites: materials, manufacturing
[20] Karki BB, Stixrude L, Clark SJ, Warren MC, Ackland GJ, Crain J. J and design. Princeton (NJ): Marcel Decker Inc.; 1988.
Am Mineral 1997;82:635. [29] Chung DH, Buessem WR. In: Vahldiek FW, Mersol SA, editors.
[21] Karki BB, Stixrude L, Wentzcovitch RM. Rev Geophys 2001;39:507. Anisotropy in single crystal refractory compounds. New York: Ple-
[22] Blaha P, Schwarz K, Luitz J. WIEN2k, Vienna University of num Press; 1968. p. 217.
Technology, 2000, improved and updated version of the original [30] Ravindran P, Fast L, Korzhavyi PA, Johansson B, Wills J, Eriksson
copyrighted WIEN code, published by Blaha P, Schwarz K, Sorantin O. J Appl Phys 1998;84:4891.
P, Trickey SB. Comput Phys Commun 1990;59:399. [31] Nye JF. Physical properties of crystals. Oxford: Oxford University
[23] Wallace DC. Thermodynamics of crystals. New York: Wiley; 1972. Press; 1985.
[24] Voigt W. Lehrbook der kristallphysik. Leipsig: Teubner; 1928. p. [32] Vajeeston P, Ravindran P, Ravi C, Asokamani R. Phys Rev B
962. 2001;63:045115.
[25] Reuss AZ. Angew Math Mech 1929;9:49. [33] Mizuno M, Tanaka I, Adachi H. Phys Rev B 1999;59:15033.

You might also like