M07 Computing Elastic Constants

Manual 07 for MD++
Calculating Elastic Constants

Keonwook Kang and Wei Cai

Nov 28 , 2007

---

Elastic constants are physical properties of crystals to relate the mechanical response to the material deformation (i.e. stress and strain) within elastic regime. To calculate the elastic constants is one simple way of validating a potential model. There are two methods to compute elastic constants: one is to monitor (Virial) stress and the other is to monitor strain energy as a function of the applied strain. Conceptually, the two approaches are equivalent except that stress is a linear function of strain and the strain energy is a quadratic function of strain. In this manual, both methods will be explained to get three independent elastic constants($C_{11}$, $C_{12}$ and $C_{44}$) for the cubic crystal materials, e.g. silicon.

---

Using Virial Stress

If Virial stress is available to compute, the task of obtaining elastic constants is relatively easy. In crystals with cubic symmetry, there are three independent elastic constants: $C_{11}$, $C_{12}$ and $C_{44}$. $C_{11}$ is the proportional constant between the stress $\sigma _{11}$ and the strain $\epsilon _{11}$, and $C_{12}$ is the proportional constant between the stress $\sigma _{22}$ and the strain $\epsilon _{11}$. We can get $C_{11}$ and $C_{12}$ from one simulation by applying strain $\epsilon _{11}$ and monitoring two components of stress $\sigma _{11}$ and $\sigma _{22}$. MD++ script si_elastic.script is given below.

# -*-shell-script-*-
# Calculate Elastic Constants of Si
# setnolog
setoverwrite
dirname = ~/Codes/MD++/runs/si_elastconst
#--------------------------------------------
#Create Perfect Lattice Configuration
crystalstructure = diamond-cubic
latticeconst = 5.430953e+00 #(A) for Si
latticesize = [     1    0    0     3
                    0    1    0     3
                    0    0    1     3]
makecrystal
#-------------------------------------------------------------
#Conjugate-Gradient relaxation
conj_ftol = 1e-4 conj_itmax = 1000 conj_fevalmax = 1000
conj_fixbox = 1
eval relax eval finalcnfile = relaxed.cn writecn
#-------------------------------------------------------------
conj_fixbox = 1 #conj_monitor = 1 conj_summary = 1
input = [ 1 1 0.001 ] shiftbox relax eval
input = [ 1 1 9.99000999000999e-04 ] shiftbox relax eval
input = [ 1 1 9.98003992015968e-04 ] shiftbox relax eval
input = [ 1 1 9.97008973080758e-04 ] shiftbox relax eval
input = [ 1 1 9.96015936254980e-04 ] shiftbox relax eval
quit

You can run the script by typing

$ bin/sw gpp scripts/si elastic.script

The above script first generates a 3 × 3 × 3 perfect diamond cubic structure of Si. It then applies tensile strain in increments of 0.1% along [100] direction by the command shiftbox, and calculate the corresponding Virial stress at each strained state by the command eval. From the log file, the stress components $\sigma _{11}$ and $\sigma _{22}$ can be obtained and plotted as a function of the applied strain $\epsilon _{11}$. Fig. 1(a) plots the result from running the script with Stillinger-Weber(SW) potential. The slopes of these lines give $C_{11}=$ 160.6 (GPa) and $C_{12}=$ 80.5 (GPa).

To compute $C_{44}$, which is the proportional constant between $\sigma _{12}$ and $\epsilon _{12}$, we can simply insert the following lines before quit in the above script.

incnfile = relaxed.cn readcn
input = [ 1 2 0.001 ] shiftbox relax eval
input = [ 1 2 0.001 ] shiftbox relax eval
input = [ 1 2 0.001 ] shiftbox relax eval
input = [ 1 2 0.001 ] shiftbox relax eval
input = [ 1 2 0.001 ] shiftbox relax eval

These commands shear the crystal in increments of $\epsilon _{12}=$ 0.1%. Again, from the log file, the stress component $\sigma _{12}$ can be obtained and plotted as a function of the applied strain. The result from SW potential is plotted in Fig. 1(b), from which we obtain $C_{44}=$ 60.2(GPa). The values for C11 , C12 , and C44 obtained here are consistent with the original paper.

• Fig.1
• 

  (a)

• 

  (b)

Using the potential energy

Sometimes, the simulation code does not implement the Virial stress calculation. In ab initio simulations, the Virial stress is usually not very accurate. In these cases, we can compute the elastic constants by monitoring the potential energy of the system, because the strain energy density $W$ is defined as

$W={\frac {1}{2}}\sigma _{ij}\epsilon _{ij}={\frac {1}{2}}\epsilon _{ij}C_{ijkl}\epsilon _{kl}$.

If the only nonzero strain is $\epsilon _{11}$, then the streain energy becomes $W=C_{11}\epsilon ^{2}/2$. We can determine $C_{11}$ by fitting $W(\epsilon _{11})$ to a parabola as shown in Fig.2(a). The elastic constnat $C_{44}$ can be determined in a similar way as shown in Fig.2(b). The elastic constants obtained in this way are $C_{11}=$160.5(GPa) and $C_{44}=$60.2(GPa), which are almost the same as the results obtained by using Virial stress.

The elastic constant $C_{12}$ is easily determined by the elastic relation

$B={\frac {1}{2}}(C_{11}+2C_{12})$

where $B$ is the bulk modulus, which we have already calculated in the manual 03. Since the bulk modulus of silicon is 108.3 (GPa) with SW potential, the elastic constant $C_{12}$ is

$C_{12}={\frac {1}{2}}(3B-C_{11})=82.2$ (GPa)

• Fig.2
• 

  (a)

• 

  (b)