Re: AMBER: About the problem of the PMEMD and cutoff

From: Robert Duke <>
Date: Mon, 15 Jan 2007 08:19:26 -0500

Yes, absolutely; if the cutoff had to be exactly half the pbc box size, for a box of any size at all that would be horrendously inefficient and completely unnecessary. The practical implications of this are simply that really really small boxes below roughly 20 angstrom on each side cannot be used (the "cutoff" of consequence is the one used to make the pairlist, so that is cut + skinnb; skinnb is 1 on uniprocessor systems and 2 on parallel systems by default for pmemd; you can reduce both of these values but not to 0, and it will cost you in performance; the default cut value is 8, and that is probably about as small as you can go). A box that size is probably roughly 800 atoms, so for explicit solvent simulations that is approaching the lower limit. I would guess that folks rarely run simulations that are not at least 10x that size, but it is not like I know what everyone is doing :-)
Regards - Bob Duke
  ----- Original Message -----
  From: 方磊
  Sent: Monday, January 15, 2007 7:45 AM
  Subject: Re: AMBER: About the problem of the PMEMD and cutoff

    Thank you for your answer,so there is no need to restrain the radius of the water box two times larger than the value of cutoff,

    The cutoff radius less than the radius of the pdb box will be ok!

  On 1/15/07, Robert Duke < > wrote:
    More precisely, the cutoff radius must be less than the radius of the maximum sphere that can be contained within the unit cell (or pbc box). The reason is simple. Any atom-atom electrostatic interaction is represented by only one atom-atom pair in the code if the atom pair formed is separated by less than the cutoff radius. Under periodic boundary conditions, there are logically an infinite number of image representations for each atom, and under the typical "minimum image convention", the atom-atom pair chosen for "direct space" calculation is the image pair with minimum separation. Now, if you allow a cutoff radius greater than the maximum sphere that fits in the unit cell, it is possible that two or more images of a given atom will fall within the cutoff. These additional images would incorrectly not be accounted for in the direct! space calculation. With pme, as long as the restriction on cutoff length is observed, all the other images are accounted for in reciprocal space. This really is not just a pmemd question; the condition applies to sander too.
    Regards - Bob Duke
      ----- Original Message -----
      From: 方磊
      Sent: Sunday, January 14, 2007 10:56 PM
      Subject: AMBER: About the problem of the PMEMD and cutoff

      Dear all,
         I heard from somebody that if the process of md will usi pme,the radius of the water box should be two times larger than the value of cutoff.
        So what does it mean? can anyone tell me?

        Thank you!

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Wed Jan 17 2007 - 06:07:42 PST
Custom Search