[Pw_forum] Re:error in band structure calculation

[Paolo Giannozzi]

On Apr 3, 2007, at 16:31 , Xinxin Yao wrote:

> When calculation="nscf", I use the initial atom position of vc-relax

"nscf" (or "bands") means "perform a non-selfconsistent calculation,
starting from a previously calculated charge density, with the atomic
positions specified in input". It is your responsibility to use atomic
positions that are consisten with the charge density. This is not the
reason for the error but it is important to get a correct band  
structure.

> 1)Why does the program delete the file "data-file.xml" when doing band
> structure calculation?

to ensure consistency of the output data. I don't remember the exact
reason but there is one.

> 2)Whether I use "bands" or "nscf", and whether I use the initial atom
> position or the relax one when doing band calculation, I always meet
> the error "starting and expected charges differ".

also for the examples? in any event, install the 3.2.1 patch: it  
corrects
a missing initialization that could lead to serious trouble in some  
cases

> I delete the file "charge-density.xml" manually, and get
> message as follows:

>      Cannot read rho : file not found
>      Initial potential from superposition of free atoms

notice the messages

>      Check: negative starting charge=(component1):  -12.900563
>      Check: negative starting charge=(component2):  -12.927055

-26 electrons? there is something wrong in the input data (or in the
pseudopotential files). One thing that doesn't seem good is the
value of "starting_magnetization" (+3 and -3) : this should be
between +1 and -1 (it is not written in the documentation but
it should be). Maybe this is the reason for so much negative
charge.

Paolo
---
Paolo Giannozzi, Democritos and University of Udine, Italy