AMBER Archive (2004)

Subject: Re: AMBER: Dielectric Constant

From: Bill Ross (
Date: Sat Dec 04 2004 - 14:22:33 CST

The higher the dielectric constant, the higher the screening of
electrostatic interactions. The distant-dependant dielectric
multiplies the constant with the distance itself.

> if lets say i want to roughly estimate the solvent
> effects, i should be using distance-dependent dielectric
> constant (though it wont be that correct) (if the simulation
> was carried out in vacuo),


> and in amber, usually it is done by turning on eedmeth=5 in the
> &ewald list, and use dielectric more than 1 (because dielc=1 is
> meant for vacum) , so there are some suggestions to use dielc in
> the range of 2-4 coz it will performed better (but I have no idea why,
> still need to do more reading on this) . am i correct?

I am not an expert on which constant to use, but I believe
the recommendations may be based simply on people's observations
of what has kept their molecules closest to solvated conformation.

Thinking out loud about the meaning, a dielectric constant > 1 with
distant-dependant mode is mainly going to affect short-range interactions
on the order of the constant chosen (either in Angstroms or amber's native
distance unit, I'm not sure which). Remember that the newer force fields
are derived for a solvent situation, which means more polar charges if I
remember right. So constant > 1 damps that bias.

> My first question is:
> Is it okay to use dielectric=80 or will it introduce some instability to the
> system?

It would reduce the electrostatic component of the force field.
Which could destabilize or overly stabilize the system depending
on what the contribution of that component is.

It would be interesting to run dynamics without electrostatics
to see what the end case would look like.

> My second question is:
> If I wanted to do a simulation in vacuo (ntb=0, igb=0), and I want to use
> distance dependent dielectric instead of constant dielectric, so I should
> turn on eedmeth=5, dielc=4 or 80 or other suggested values, and is it okay
> to use cut=16? Or should I use larger value?

The larger your constant, the less effect a short cutoff will have.

The longer the simulation, the greater the effect of long-range
interactions - being weak, they need time to make a difference.

My guess is that simulating solvent with dielectric is such
an approximation that few if any people attempt long simulations
with any expectations of accuracy, aside from getting impressions
of what unfolding paths might be available and the like.

> My third question is:
> Which is better to use? constant dielectric (eedmeth=4) or distance
> dependent dielectric(eedmeth=5) for simulation in vacuo? Because from what I
> read, simulation using the distance dependent dielc could make the protein
> to shrink as a result of the columbic interactions that are not properly
> shielded.

I think the best thing would be to try both approaches, since you
will then know what the effects are for your system. Vacuum simulations
run so quickly, you may be able to find if the protein shrinks in a
few minutes of run time.

If you do this, I hope you will report your conclusions here.

This is the sort of thing that would make an interesting teaching-the-
basics demo (which my crown ether demo tries to be). It could be done
for proteins and nucleic acids, possibly finding different values for each.

> I am so confused, and I would appreciate it if somebody could give me some
> hints or suggestions or corrections. And I would appreciate it very much if
> any of you could suggest a few good reading materials on this area ( I have
> read some but I need to read more, maybe there are some good articles that
> im not aware of).

There must be something in the FAQ. Presumably you have seen

But no list of basic books..
Allen & Tildesley (sp?) is a basic reference on the methods, and
there should be more somewhere in the manual.

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to