Guangping, There is a difference between the broadening width in the DOS/PDOS calculated with the ProjectedDensityOfStates block or eig2dos, and the one you get through interactions of your molecule with some other system (electrodes, surfaces, clusters, other molecules, whatever). In the first case, which is what you want to know, it has nothing to do with being an open or closed system. Rather, the broadening simulates the effect of a finite temperature on the DOS. if your k-point mesh is sufficiently dense, the role of the gaussian broadening is to smoothen the DOS. Use physical wisdom to estimate an initial value for the gaussian broadening: how much is $k_BT$ at room temperature? Starting from that value, increase and decrease it value by small amounts: do you see any features that become less evident as you change the broadening? My rule of thumb is: the "good" broadening is the lowest which gives you a PDOS that looks sufficiently smooth without losing any of the peaks or important features. Of course, this is quite subjective, and changes from system to system. In other words, you have some flexibility in using the broadening - use it sensibly.
Marcos On Mon, Apr 12, 2010 at 1:59 PM, Guangping Zhang <[email protected]> wrote: > Hi,Marcos, > "peak width" I said means peak width (in eV) for broadening (gaussian or > lorentzian).Since I get the EIG file,I want to get the DOS ,I must to set > the broadenin width,So how to define it,for different values have very > different DOS results.In transiesta,the DOS give automatically.In that > situation,the system is open ,so there is an interaction between the > melecular and the electorde,so there is a broadenin width from the > interaction energy.But if I calculate a isolated molecular,how it comes the > broadenin width? > Am I right ? > Best wishes! > >
