Definitions
In CRYSTAL, the effect of strain is defined as follows:
- The cartesian axes remain unchanged
- The cartesian components of a vector r transform as follows:
In particular, if A is the matrix containing the cartesian components of the three elementary
translation vectors, a
1,
a
2
and a
3
The strained matrix, A' is:
A' = (1+s) A
The symmetry of the elastic tensor and Voigt notation
The tensor [sJ can be decomposed into two parts:
- an antisymmetric part, [[J, that corresponds to a pure rotation
- a symmetric part, [sJ, that corresponds to a pure strain (without rotational components)
Then, if we are interested only in the strain part of the transformation, we can refer to the
following pure strain tensor:
As only six components of the symmetric matrix are independent, a more compact one-index
notation can be adopted.
Let us define, according to Voigt's convention:
In this convention the strain tensor becomes:
and can be expressed as a vector:
The ZnO case
To illustrate the effect of strain, we will use the ZnO wurtzite structure (hexagonal 6mm class).
The ZnO cell parameters are:
a=3.2859 c=5.2410
The three direct lattice vectors are labelled a
1
, a
2
, a
3
; E F K are, as usual, the angles between a
2
and a
3
, a
1
and a
3
, a
1
and a
2
, respectively. The convention adopted in CRYSTAL for the a
1
and a
2
orientation is given in figure 1. The cartesian components of a
1
and a
2
are reported. The A
matrix, is given on the right.
Figure 1: Hexagonal cell of wurtzite in the (xy) plane, according to CRYSTAL convention.
As a first example, let's apply the following strain to ZnO:
The A matrix is modified by the strain into A'
= a
= a
E' = 90
0
F' = 90
0
K' = 120
0
Figure 2: Deformation of the ZnO hexagonal cell, when the strain s is
applied.
The undistorted and distorted cells are represented in light and dark
grey, respectively. The volume variation is also represented in dark
grey. The new lattice parameters are reported on the right side of the
figure.
As a second example consider:
From A', it is easily seen that = a, = c; E = F = 90
0
For and K' we get:
For small H values, we can disregard terms quadratic in H, so that we get for and K' (use a
Taylor expansion):
and or in degrees.
For H = 00l (a typical value we will use in the following), A|a
1
| = -0.43 and AK = 01o .
In summary (length in and angles in degrees; variation percentage is also given in
parentheses):
I)
a'
1
=3.2718
(-0.43)
E '= 90
0
(0.00)
a'
2
=3.2859
(0.00)
F' = 90
0
(0.00)
a'
3
=5.2410
(0.00)
K'=119.57
0
(-0.36)
We can consider now the effect of deformations II, III and IV, closely related to the one just
discussed. It turns out that II has the same effect as I, to first order, and III is simply the
superposition of I and II; IV is a pure rotation (again to the first order; actually the modulus of
the a' lattice parameter changes 2 . 10
-4
), so that the translation vectors and the angles between
them do not change. This means that we can use, alternatively, I or II or III; their effect is, to
first order, exactly the same; for example, the energy difference, AE with respect to the
unstrained case, is shown in the following figure. AE for deformation IV is null to first order.
The choice among I, II and III will be eventually dictated by symmetry considerations.
Figure 3: Variation of ZnO total energy (in mhartree) when the four
strain tensors I, II, III, IV (see text) are applied. The energy
difference A E
0,I
or A E
0,II
are 3.35 10
-4
hartree, whereas the second
order differences ( A E
0,IV
or A E
I,II
) are of the order 10
-7
, 10
-8
.
Note that A E
0,III
is four times H ( AE is proportional to (2H)
2
and
H
2
, see the elastic section below).
II)
a'
1
=3.2717
(-0.43)
E '= 90
0
(0.00)
a'
2
=3.2861
(0.00)
F' = 90
0
(0.00)
a'
3
=5.2410
(0.00)
K'=119.57
0
(-0.36)
III)
a'
1
=3.2575
(-0.86)
E '= 90
0
(0.00)
a'
2
=3.2861
(0.00)
F' = 90
0
(0.00)
a'
3
=5.2410
(0.00)
K'=119.14
0
(-0.72)
IV)
a'
1
=3.2861
(0.00)
E '= 90
0
(0.00)
a'
2
=3.2861
(0.00)
F' = 90
0
(0.00)
a'
3
=5.2410
(0.00)
K'=120
0
(0.00)
The elastic tensor
Definitions
The elastic constants are second derivatives of the energy density with respect to strain
components:
where V is the volume of the cell and |c| is a rank 4 tensor with 3 x 3 x 3 x 3 = 81 elements
(called constants).
As seen before, the strain s can be expressed with a single index (i=1,6), using Voigt's notation.
The elastic constants then become:
with i,j=1,6. It is obvious from the definition that this matrix is symmetric, so the elastic tensor
has a maximum of 21 independent elements.
The elastic tensor and the symmetry
Due to the symmetry of the crystal, some of the constants are zero, or symmetry related. In order
to know which independent components must be considered, a preliminary run (a TESTGEOM
run) must be performed using the TENSOR keyword. CRYSTAL output indicates which
constants are independent, and the coefficients relating the symmetry equivalent constants.
How to proceed
A Taylor expansion of the unit cell energy to second order in the strain components yields:
If E(0) refers to the equilibrium configuration, the first derivatives are zero. According to
equation (2), the elastic constants can then be obtained by evaluating the energy as a function of
strain.
The energy derivatives can be evaluated numerically (CRYSTAL03) or analytically
(CRYSTAL06), with respect to the cell parameters.
When the unit cell is deformed (using the ELASTIC option), the point group may be reduced to
a subgroup of the original point group (see examples below). The new point group is
automatically selected by the code. For each deformation of the unit cell, the atoms must be
allowed to relax (In some cases this relaxation is not necessary, as the atoms occupy special
positions). The optimization is performed by using the OPTGEOM option (see manual).
The ZnO case
To illustrate the calculation of an elastic tensor, we will continue to detail the ZnO wurtzite
example.
First of all, in order to know which are the independent tensor components, we run the crystal
executable with the following geometry input:
ZnO wurtzite elastic tensor
CRYSTAL
0 0 0
186
3.2859 5.2410
2
30 0.66666666666667 0.3333333333333 0.
8 0.66666666666667 0.3333333333333 0.38268
TENSOR
4
TESTGEOM
END
The wurtzite compounds have the 6mm hexagonal symmetry. The TENSOR directive produces
the following output:
THIS ROUTINE EVALUATES THE NON ZERO ELEMENTS
OF TENSORS OF PHYSICAL PROPERTIES UP TO ORDER 4
ORDER OF THE TENSOR : 4
| 1 A 1 E 1 B . . . |
| 1 A 1 B . . . |
| 1 D . . . |
| 1 C . . |
| 1 C . |
| & |
EACH LETTER CORRESPONDS TO AN INDEPENDENT COMPONENT.
INTERDEPENDENT COMPONENTS (WITH SAME LETTER) ARE RELATED BY
THE COEFFICIENT THAT PRECEDES EACH LETTER.
THE "&" SYMBOL MEANS 1/2(C11-C12) (OR 2(S11-S12)) FOR THE ORDER
4 C AND S ELASTIC TENSORS IN THE TRIGONAL OR HEXAGONAL CASES.
The number of independent terms in various tensors, in particular for the elastic and piezoelectric
tensor can be found in standard textbooks, for example in Nye[1]. These tables must be used
with some care, because the structure of the tensor depends on the adopted convention for the
orientation of the cartesian axes with respect to the unit cell vectors. The output of the TENSOR
routine might show permutation with respect to standard definitions[1])
There are only 5 independent components (5 different letters) to evaluate:
A: c
11
= c
22
B: c
13
= c
23
C: c
44
= c
55
D: c
33
E: c
12
The symbol & indicates that c
66
can be deduced from two independent components [c
66
= ~ (c
11
+
c
12
)]. This particular case happens for trigonal and hexagonal systems only.
To define the complete elastic tensor, 5 different strains must be applied. The c
ij
constants can be
separated in two groups:
- Diagonal terms. They are directly related to second derivatives of the energy.
- Off-diagonal terms. The second derivative is obtained by at least two deformations.
The diagonal terms
c
11
The [ J tensor for the calculation of c
11
is:
The energy expression is:
where a and b are the coefficients of a polynomial fit of E versus H (see examples below). Then
(see equation (2));
c
33
The [ ] tensor for the calculation of c
33
is:
The energy expression is:
Then:
c
44
The [ ] tensor for the calculation of c
44
is:
NB: In Voigt's notation
The energy expression is then:
Then
The off-diagonal terms
c
12
The [ ] tensor for the calculation of c
12
is:
The energy expression is:
The output of TENSOR shows that c
11
= c
22
(and c
12
= c
21
by definition), then
Then
(c
11
+
c
12
) = b/V
c
13
The [ J tensor for the calculation of c
13
is:
The energy expression is:
Then
(~c
11
+
~c
33
+c
13
) = b/V
Computational conditions
To obtain E as a function of a given , the energy must be computed for various H (the strain
amplitude). Particular care is required in the selection of the computational parameters and of the
points where the energy is evaluated, in order to avoid large numerical errors in the fitting
procedure.
As regards computational conditions:
- All the calculations were performed with the same "reference" geometry (the undistorted
one) for the selection of integrals, by using the FIXINDEX option.
- The point group may reduce to a subgroup. Then internal relaxation of the atoms may be
necessary. In this case, high accuracy is required in the convergence criterion of the
optimization. The following options were used to modify the default values (see
OPTGEOM option):
OPTGEOM
TOLDEG
0.00006
TOLDEX
0.00012
TOLDEE
8
ENDOPT
(default values, are less severe, namely 0.0003, 0.0012 and 7)
- Very high accuracy is also needed in the SCF step:
- TOLDEE
- 10
The default value used in CRYSTAL06 for standard SCF is 5 (see manual).
The fitting
In the example below, referring to c
33
, 15 runs have been performed (with the strain presented
above) with -0.035 < H < 0.035 and step 0.005.
The table included in figure 3 gives the 15 energy points that are fitted with polynomials of
various degrees. The value of the second derivative, b, of the fitted function at the equilibrium
geometry, is presented in Table 1 . It turns out that, in the explored interval, the first acceptable
polynomial must contain a fourth order component (H