The ABINIT Users Mailing List ( CLOSED )

[Z_Lin <applelinux@gmail.com>]

> > So if the system is excited by a laser pulse, can the matrix element be
> > the same? I mean, since the screen potential is from the fact that the
> > electrons react to contribute to the effective potential. Wouldn't this
> > effective potential change when the electrons are excited in terms of
> > non-zero electron temperatures, temperature dependent?
> Yes, but it depends which time scales you are looking at. In a first
> approximation, immediately after the pulse, the electrons are excited in a
> really unequilibrated way. Then (normally) electronic processes allow
> thermalization of the electron population and the distribution of the
> states should be fermi-like. Later, the e-p coupling kicks in and the heat
> is progressively (sometimes quite quickly!) transferred or shared with the
> lattice.

Yes. I understand that the thermalization of hot electrons is
typically fast, on the scale of tens femtoseconds whereas the
relaxation of hot electron energy to the lattice in metals is much
slower, on ps scale. The question that I intended to ask was: if we
assume the establishment of a homogenous hot electron temperature
(electrons have thermalized themselves and formed a fermi
distribution) and look at the equilibration process of electron with
lattice, can we take the matrix element from previous GS calculation
as the electron-phonon interaction at this stage? I presume one need
to calculation at non-zero fermi temperature to obtain these elements
as setting a combination of tsmear and occopt=3 in the smearing for
metals like what you mentioned below?

>
> > And here comes another questions while I was reading your previous email
> > about the physical electron temperature tphysel and the laser paper by
> > CEA group PRL 2006.  I assume that they specify the value of tphysel to
> > be 6eV in abinit?
> tphysel is probably not necessary. To my knowledge they used just a fermi
> distribution. tphysel is for intermediate physical temperatures, when the
> physical smearing is not enough to improve your kpoint convergence. Then
> you can add a second (rapidly decaying) smearing function in order to
> improve your k convergence without destroying your electronic temperature
> dependence.

> > BTW, the umklapp processes is included in the calculations of matrix 
> > element?
> hmmmm. I'm not sure I know the answer to that one. You mean for q vectors
> out of the first BZ? I'll ponder that some more. Anyone have an idea?

Yes. That is exactly what I meant. But roughly the matrix element
varies as 1/(q+G)^2. So that value might be quite small though.

Thanks.
Zhibin