Author Topic: constraint on no.of atoms in simulation  (Read 3821 times)

0 Members and 1 Guest are viewing this topic.

Offline Zexter

  • Heavy QuantumATK user
  • ***
  • Posts: 53
  • Country: in
  • Reputation: 0
    • View Profile
constraint on no.of atoms in simulation
« on: December 17, 2010, 09:19 »
Hi
    I need to simulate my design in ATK 10.8.2 the count on the atoms goes to 8000  (Carbon) and my configuration is
Red hat linux enterprise edition 5(64 bit)
8GB DDR3 memory
512MB nvidia zotac graphics card
quad core pc 2.25Ghz

I will be using only single zeta. Is this thing is possible to be done on my rig

Is it possible to compute potential energy minima in the configurations. if so what has to be done to get it.
« Last Edit: December 17, 2010, 09:25 by Zexter »

Offline Anders Blom

  • QuantumATK Staff
  • Supreme QuantumATK Wizard
  • *****
  • Posts: 5565
  • Country: dk
  • Reputation: 93
    • View Profile
    • QuantumATK at Synopsys
Re: constraint on no.of atoms in simulation
« Reply #1 on: December 17, 2010, 09:34 »
Unfortunately, 8000 atoms is a huge number for DFT; you are typically limited to below 1000 atoms, although we are working hard to push this number up (and with 8 Gb you can probably go a bit higher, esp. with carbon which has a small basis).

With the Huckel model 8000 atoms might be doable with ATK 11.2, it too early to say, we're working on some improvements in the memory footprint. Our Huckel basis sets for carbon are excellent for graphene and nanotubes, we believe the results are as good as DFT (or even better sometimes). Now, our Huckel cannot yet do geometry optimizations since there are no forces, but soon ATK will have classical potentials too, in particular for carbon, and that is what I think you really need to find the energetic minimum you are looking for, in such large structures.

Note that the SingleZeta basis set is for testing only; it's not a basis set that can be trusted to deliver accurate results, even for carbon. You need at minimum SingleZetaPolarized or DoubleZeta.

Offline Zexter

  • Heavy QuantumATK user
  • ***
  • Posts: 53
  • Country: in
  • Reputation: 0
    • View Profile
Re: constraint on no.of atoms in simulation
« Reply #2 on: December 17, 2010, 10:03 »
Hi
whether the huckel basis set calculation can relieve me my computation burden
whether my simulation with its 8000 atom is possible with huckel basis set I nvnl 10.8.2 which I presently have.

Offline Anders Blom

  • QuantumATK Staff
  • Supreme QuantumATK Wizard
  • *****
  • Posts: 5565
  • Country: dk
  • Reputation: 93
    • View Profile
    • QuantumATK at Synopsys
Re: constraint on no.of atoms in simulation
« Reply #3 on: December 17, 2010, 10:33 »
Well, Huckel uses about the same amount of memory as SingleZeta, actually. However, this depends on the basis set (the Cerda basis sets are larger). There is however a large benefit on the mesh, these are smaller for Huckel, so there's a memory reduction. I think 8,000 atoms in not realistic right now, with either method, actually. At least not with 8 Gb, and certainly not in ATK 10.8.