Hello Users,
I would like to determine the fastest bond frequency so that I can choose the appropriate time step. I am using a harmonic potential for the bond and the force constant units are kcal/(mol*angstroms^2). If my physics has served me well, then I believe the vibrational frequency can be determined by the following equation:
w = sqrt(k/m) where w is the frequency, k is the spring constant and m is the reduced mass of the two bonded atoms.
In order for the equation to work, it seems that k should be in units of newtons/meter. I’m lost as to how to convert kcal/(mol*angstroms^2) to N/m. Is there a better way to calculate the vibrational frequency?
Thanks,
Casey
Hello Users,
I would like to determine the fastest bond frequency so that I can choose
the appropriate time step. I am using a harmonic potential for the bond and
that would only work for simple gas phase system.
why don't you just determine the necessary time
step empirically?
the force constant units are kcal/(mol*angstroms^2). If my physics has
served me well, then I believe the vibrational frequency can be determined
by the following equation:
w = sqrt(k/m) where w is the frequency, k is the spring constant and m is
the reduced mass of the two bonded atoms.
In order for the equation to work, it seems that k should be in units of
newtons/meter. I'm lost as to how to convert kcal/(mol*angstroms^2) to N/m.
Is there a better way to calculate the vibrational frequency?
you are disregarding the fact that frequencies appearing
may not only be in the lowest excited mode.
the time step requirements also depend on the steepness
of the repulsive part of non-bonded potentials.
as a rule of the thumb you can take the following guidelines.
molecular systems at 300K with flexible bonds to hydrogen atoms
typically require a time step of 0.25fs seconds as a conservative
choice (people sometimes push the limits by going to 0.5fs).
you can raise the time steps roughly with sqrt(mass) if you
go to heavier elements. but you also need to account for
higher velocities as elevated temperatures by lowering the
time step. if you use rigid objects or shake things get different
again. it is fairly easy though to determine a good time step
empirically with a set of test calculations of a small system.
cheers,
axel.
I'll actually disagree with Axel for once. Finding
the stiffest spring in your system and calculating
it's frequency, which is just a function of mass and
spring constant, as in your formula, is a fine
way to estimate a timestep. The frequency is
temperature independent (to first order) and the
system will never oscillate faster than that under
normal conditions. Doesn't matter if it's a gas or
liquid or solid. A timestep of around 1/10 to 1/20
of that period is often fine to track the motion
accurately. I believe the C-H
period is about 10 fmsec if I remember and that
is often the stiffest bond in organic systems.
Hence a timestep of 0.5 to 1.0 fmsec. All you
have to do is convert units in your formula. Just
work thru the math.
Steve