[Pw_forum] Further discussion on electronic DOS calculation

yanmingma at sohu.com yanmingma at sohu.com
Sat Aug 13 16:17:03 CEST 2005


Dear Stefano de Gironcoli and all users<p><p>As I understand, the electronic DOS calculation in pwscf as illustrated in example8 is ok for the semiconductor. <p>However, to deal with metals, we might meet some problems. In this case, we need very accurate description of the Fermi energy for metals.<p>In the SCF calculation, we usually use a Gaussian broadening method to obtain the Fermi energy.<p>Then in the DOS calcualtion, we use a more accurate tetrahedral technique to get the DOS curve.<p>But problems come due to the somewhat mismatch in the two steps, particular when one is seeking for the accuate electronic DOS at Fermi energy (N(0)). In principle, if we integrate the DOS below Fermi energy, then the integration should be equal to the total valence electron numbers. Unfortunately, in most cases I tested, the two numbers are different due to the obvious mismatch in the two steps. So I can not very trust the results of N(0) by the above procedure, even I know the N(0) is very close to the real. <p>Then I tried to do the following.<p>(1)I use a Gaussian broadening method to obtain the Fermi energy. then I use Gaussian broadening method to obtain the DOS curve. In this case, the integration get a very good agreement with the total electrons. However, due to the limitation in calculation of DOS in Gaussian broadening method, the calculated N(0) is not very trustable.<p>(2)I tried to use tetrahedral method to find the Fermi energy in SCF step. But I failed to get a reasonable Fermi energy if I remember correctly.<p><p>So, in a metal, how to get the very accurate N(0) is quite bothered me using PWSCF. Could anyone describe me how to do this?<p>I will highly appreciate your answer.<p><p>Regards<p><p>Yanming <p>  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20050813/2d9770fe/attachment.html>


More information about the users mailing list