Suman, I think the answer you're looking for is 42. Only 42-atom simulations are considered by peer review ;)
But joking aside, you should look to previously-published ab-initio results in your system of interest or something close to it. Even if such results exist, you should perform methodological checks on system size, i.e. size-convergence simulations with respect to: the number of layers in a surface slab, or supercell dimensions if you're simulating a bulk compound (perhaps with a defect or other non-homogeneity), different sizes of nanowire, etc. Some of the most important current problems in nanoscience involve quantum size-effects, and you don't really know how many atoms a "good" simulation requires until you try it- the physics of a particular solid-state system can vary dramatically based on small size increments! So, to answer the question you posed, no real "number of atoms" guideline exists. Use what is feasible and timely for your computational setup while giving trustworthy, meaningful, repeatable results! By the way, the "T" in siesta stands for "Thousands" (but this only really makes sense with the order-N functional). The number of atoms in a SIESTA simulation could be anywhere in the range {10^0, 10^4}. -- *Abraham Hmiel* Katherine Belz Groves Fellow in Nanoscience Xue Group, SUNY College of Nanoscale Science and Engineering http://abehmiel.net/about On Fri, Jan 24, 2014 at 8:21 AM, Suman Chowdhury <sumanchowdhur...@gmail.com > wrote: > Dear all, > what is the typical number of atoms one should use for publishable results > using SIESTA? >