[Pw_forum] Re: on finite electric methods to calulate vibrational properties

[Paolo Umari]

>Dear Stefano Baroni ,
>
>  I'm sorry for my spelling error.
>   I want to investigate Raman spectra in the case of large   disordered 
>system from first principles and use the finite
>electric field methods.
>The dielectric polarizability tensor can be calculated by taking dinite 
>differences of  atomic forces.

If you want to calculate the raman tensor you need either the 
first derivative
of the dielectric polarizability tensor with respect to the atomic 
displacements or the second derivative of the atomic forces with
respect to twice the electric field.
The second choice is the best one with the finite electric field method

>There are several questions to ask you.
>  1. We obtained the diagonal terms, by considering values for the 
>electric field of
>-2h,-h, 0, h and 2h in the cp.x code[tefield=.true.
>                                                         tcg = .true.,
>                                                     epol = 1 or 2 or 3
>                                                    efield = -2h,-h, 0, h 
>or 2h],
>and by using the five-point formular . Is it right?

It's right, but I should check the convergence by using only the 
three-point formula

>  2. For the calculation of the off-diagonal terms, two different 
>electric 
>fields
>are required. This need apply electric fields of intensity 
>-2h/1.414,-h/1.414,
>0, h/1.414 and 2h/1.414 along the three different couples of cartesian 
>directions, and apply eauation.
>But how can I apply simultaneously finite fields along two different 
>Cartesian directions
>in cp.x code?[ESSPRESSO-3.2]

As the Berry's phase implementation  provides  electric fields along the 
three carthesian  directions, to consider an electric field along an 
arbitrary direction you  should consider its components along the
three cartesian direction.
By now, 2 components are implemented, the second one is defined by
the input variables: tefield2,epol2,efield2 (defined as 
tefield,epol,efield)

Note that only the options epol=3 and epol2=3 are now available in case 
of parallel computing

>3. In this way, we could obtain all the dielectric polarizability tensor. 
>Then the
>Raman susceptibilities associated to the normal mode n can be obtianed 
>according to the equation.
>How to implement it ? By a small code?

For the Raman tensor, you should consider the second derivatives of the 
atomic forces....


Best wishes,

PaoloU


>  Best!
>  I'm counting on the new version.