[Pw_forum] Pressure-induced phase transition

[Hongsheng Zhao]

On 07/19/2011 12:45 AM, Amin Torabi wrote:
> Dear Hongsheng,
>
> I don't understand why I should use relax instead of vc-relax. As far as
> I understand, when a system is under pressure, both the atomic positions
> inside a unit cell, as well as the dimensions of the unit cell changes.
> So, why should I keep the unit cell parameters fixed if I am interested
> in finding phase transitions?

In fact, you can do relax or  vc-relax, it depend on the actual ways 
used by you to study the   Pressure-induced phase transition.  Let me 
give you a more detailed description as follows:

1-  If you only use one type of initial phase to do this type of study. 
  Then you should use the vc-relax when changing the external pressure. 
  And then see what's the ultimate unit cell looks like for each 
pressure step.  For convenience, you can also plot the 
binding-energy/formation-energy vs external-pressure to see whether 
there are some abrupt change points - of course, there are many other 
physical properties you can consider to assist you judging the first 
order phase transition, I just give you a example here - on the plot.

2-  If you have two or more  initial phases and want to find the phase 
transition pressure among them.  In this case, you can do the relax 
calculation for each phase  by impose external pressure in certain 
range.  And then you can plot the E-V or H-V plots for each phase in the 
same Cartesian system  to find the  phase transition point.

Basically, the first method (vc-relax) enable you to find or predict 
some possible phase you don't known in advance.  The second method 
(relax) enable you find the phase transition pressure for known phases.

But, even by using the vc-relax, I don't think it can be served as a 
perfect method for new-phase finding an searching technology.  This is 
because the vc-relax just changing the unit cells ( cell parameters and 
atoms positions ) in a series of localized positions.  So I don't think 
by using this method you can do the job like universal structure 
prediction just as the uspex or similar codes done for you.

Regards
>
> Thanks


-- 
Hongsheng Zhao <zhaohscas at yahoo.com.cn>
School of Physics and Electrical Information Science,
Ningxia University, Yinchuan 750021, China