Environmental Engineering Reference
In-Depth Information
Figure 1.7
Normalized velocity autocorrelation function with different
number of layers (NL) of Nose-Hover and Langevin heat bath applied.
Exceptforthesolidline,theotherthreelinesalmostoverlapwitheachother.
velocity-scaling type heat bath, withthe scaling factor [38]
1
T
0
T
−
1
1
/
2
+
t
τ
T
γ
=
,
(1.35)
where
t
is the time step,
τ
T
is the relaxation time,
T
0
and
T
are the aimed temperature and instantaneous temperature,
respectively. The relaxation time
τ
T
should be properly chosen to
avoidunrealisticallylow-temperaturefluctuationswithsmall
τ
T
,and
the inactive sampling with large
τ
T
(e.g.,
τ
T
→∞
) [39]. Based on
such considerations, we set
τ
T
=
t
×
10
4
(8 ps) in our study.
As shown in Fig. 1.8, with Berendsen heat bath, the NW fails to
reach the aimed temperature (310 K and 290 K) at two ends.
More seriously, large TJ persists regardless of the number of heat
bath layers applied. This even poorer performance of Berendsen
heat bath is a consequence of its inability to reproduce canonical
ensemble, because of the artificial velocity-scaling scheme [39],
which has been reported to cause artifacts in various studies