Escolar Documentos
Profissional Documentos
Cultura Documentos
analysis of MD results
1.
2.
3.
A(t)
Time, t
Although the value of A(t) is changing randomly, for two measurements
taken at times t and t that are close to each other there are good
chances that A(t) and A(t) have similar values => their values are
correlated.
If two measurements taken at times t and t that are far apart, there is
no relationship between values of A(t) and A(t) => their values are
uncorrelated.
The level of correlation plotted against time would start at some value
and then decay to a low value.
tcorr
A(t)
Time, t
and multiply the values in new data set to the values of the original data set.
Now let us average over the whole time range and get a single number
G(tcorr). If the the two data sets are lined up, the peaks and troughs are
aligned and we will obtain a big value from this multiply-and-integrate
operation. As we increase tcorr the G(tcorr) declines to a constant value.
The operation of multiplying two curves together and integrating them over
the x-axis is called an overlap integral, since it gives a big value if the curves
both have high and low values in the same places.
The overlap integral is also called the Correlation Function G(tcorr) =
<A(t0)A(t0+tcorr)>. This is not a function of time (since we already integrated
over time), this is function of the shift in time or correlation time tcorr.
Velocity
Below is a time dependence of the velocity of a H atom on a Hterminated diamond surface recorded along the MD trajectory.
25
50
75
100
T im e fs
Time,
N
r
r
r
1
1 maxr
G ( ) = v i (t 0 ) v i (t 0 + ) i, t =
v i (t 0 ) v i (t 0 + )
0
N i =1 t max t =1
0
<v(t0) v(t0+t)>
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
Timefs
Time,
The resulting correlation function has the same period of oscillations as
the velocities. The decay of the correlation function reflects the decay in
the correlations in atomic motion along the trajectories of the atoms, not
decay in the amplitude!
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
I( ) =
exp(2i)G()d
Spectral peaks corresponding to the specific CH stretching modes on the Hterminated {111} and {100} diamond surfaces.
L. V. Zhigilei, D. Srivastava, B. J. Garrison, Surface Science 374, 333 (1997).
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
600
Velocity, m/s
400
200
0
-200
-400
-600
0
5000
10000
Time, fs
<v(t0) v(t0+t)>
500
1000
Time, fs
1500
2000
350
300
250
200
150
100
50
0
25
50
75
100
Frequency, cm-1
r
r
r r
C( r ) = ( ri ) ( ri + r )
and
r
where ( ri ) =
r r
(ri r j ) = 1
r r
( ri + r ) =
r r r
(ri + r r j ) =
r r
(r rij )
Therefore
N
r
r
r r
C( r ) = ( ri ) ( ri + r ) i =
r r
(r rij )
1
=
N
r r
(r rij )
r
r
C( r )
V
c( r ) =
=
0
N2
r r
(r rij )
r
c( r ) =
1
N0
r r
V
(r rij ) =
N2
r r
(r rij )
For isotropic system g(r) can be averaged over angles and calculated
from MD data by calculating an average number of particles at distances
r
r r+r from any given particle:
N (r) = c( r )
r
angle
g(r) = N r /(4 r r)
Thus, we can define the pair distribution function, which is a realspace representation of correlations in atomic positions:
1
g(r) =
4Nr 2 0
1
(r rij ) =
2
2
Nr
0
j =1 i =1
i j
(r r )
j =1 i > j
ij
g(r) can be calculated up to the distance rg that should not be longer than
the half of the size of the computational cell.
Although g(r) has the same double sum as in the force subroutine, the
range rg is commonly greater than the potential cutoff rc and g(r) is
usually calculated separately from forces.
Calculation of g(r) for a particular type of atom in a system can give
information on the chemical ordering in a complex system.
g(r) for pure Ni in a liquid state at 1773 K (46 K above Tm). Experimental
data points are obtained by Fourier transform of the experimental structure
factor.
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
1
g(r) =
4Nr 2 0
1
(r rij ) =
2
2
Nr
0
j =1 i =1
i j
(r r )
j =1 i > j
~
g (r) = g(r) - 1
Reduced PDF:
G(r) = 4r0 ~
g (r)
H(r) = 4r 2 0 ~
g (r)
or
15
G(r) (Atoms/ )
10
-5
-10
0
12
16
20
Distance r ()
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
ij
R(r) = 4r 2 0 g(r)
Radial Distribution Function
R(r) for FCC and liquid Au
24
6
12
n =
r2
R ( r )dr
r1
N
n =
V
r2
g ( r ) 4 r 2 dr
r1
Source
Sample
Q =| k f ki |=
ki
4 sin
S (Q) I (Q)
kf
Detector
1
g (r ) = 1 + 2
Q[ S (Q) 1] sin(Qr )dQ
2 r 0 0
0 is average number density of the samples
Q is the scattering vector
[1]
()
i =1
r
r
where Q is the scattering vector, ri is the position of atom i
fi is the x-ray or electron atomic scattering form factor for the ith atom.
r
The magnitude of Q is given by Q = 4 sin / , where is half the
angle between the incident and scattered wave vectors, and is the
wavelength of the incident wave.
I (Q) = f i f j
sin(Qrij )
Qrij
j =1 i =1
f
i =1
2
i
I (Q)
N
f
i =1
2
i
=1+
fi f j
f
i =1
sin(Qrij )
2 j =1 i < j
Qrij
2 N N sin(Qrij )
S (Q) = 1 +
N j =1 i < j Qrij
[2]
g(r) =
2Nr 0
(r r )
j =1 i > j
ij
The expression for the structure function can be now reduced to the
integration over r (equivalent to Fourier of g(r) ):
S (Q) = 1 + 4r 2 0 g (r )
0
sin(Qr )
dr
Qr
[3]
damping function W(r) can be used to replace the sharp step function at
Rmax by a smoothly decreasing contribution from the density function at
large interatomic distances and approaching zero at Rmax:
r
Rmax
sin
sin(Qr )
R
2
S (Q) = 1 + 4r (r )
W (r )dr
max
W (r ) =
Qr
0
r
Rmax
25
(1 1 1)
(3 1 1)
(2 0 0)
20
(3 3 1)
10
(4 2 0)
(4 2 2)
(2 2 2)
(4 0 0)
Rmax,
40.0
80.0
100.0
140.0
20
15
with W(r)
(2 2 0)
without W(r)
with W(r)
(1 1 1)
25
(3 1 1)
(2 0 0)
15
(2 2 0)
10
(3 3 1)
(4 2 0)
(4 2 2)
(2 2 2)
5
(4 0 0)
-5
0
2
-1
-1
Q ( )
Q ( )
Competition between
heterogeneous and
homogeneous melting
Phys. Rev. B 73, 184113 2006
Simulation:
Z. Lin and L. V. Zhigilei,
J. Phys.: Conf. Series 59,
11, 2007.
Shifts of diffraction
peaks reflect transient
uniaxial thermoelastic
deformations of the film
an opportunity for
experimentally
probing ultrafast
deformations.
H. Park et al. Phys. Rev. B 72, 100301, 2005.
University of Virginia, MSE 4270/6270: Introduction to Atomistic Simulations, Leonid Zhigilei
505
746
964
1137
1262
1338
(1 1 1)
(2 2 0)
(3 1 1)
uniaxial lattice
deformation
Normalized Peak Height
1327
0.8
homogeneous
melting
0.6
0.4
uniform lattice
heating
0.2
0
0
10
12
14
Time (ps)
(a) 0 ps
(b) 20 ps
(d) 150 ps
(e) 200 ps
(c) 50 ps
(f) 500 ps
Evolution of diffraction
profiles can be related to
time-resolved pumpprobe experiments, with
connections made to the
microstructure of the
target.
P.E. =
1
N
N N
i =1
r
N 2
U(| rij |) = 2
r U(r )g(r )dr
V
j>i
For pairwise interaction, g(r) can be used to calculate the pressure in the
system,
P=
Nk B T 2 N
dU ( r ) 3
r g ( r ) dr
2
V
dr
3V
0