October 2010
Hi Everyone,
Since Alison is travelling at the moment, I'm moonlighting as the
newsletter coordinator for October. It been a busy month since Stefano
di Gironcoli has been putting together the project results from the
school.
All the best!
Sinead Griffin
:: CONTENTS ::
1) Jobs
2) Journal Access from ICTP
3) High Performance Computing School
4) Article of the month: Ab Initio Random Structure Searching
5) Group-project update
:: JOBS ::
1) The European Psi-k network keeps a mailing list with an average of 1-2
announcements per day, mostly jobs in electronic structure but also conferences
etc. In order to receive the announcements by email, one needs to register at
the the Psi-k portal http://cselnx9.dl.ac.uk:8080/portal
Registration is free and open to all.
2) The list of opportunities in Condensed Matter Physics at ICTP (including
postdoctoral fellowships and short visits) can be found at
http://cmsp.ictp.it/job-visiting-opportunities
:: JOURNAL ACCESS FROM ICTP ::
ICTP maintains, through special agreements with publishers, a free delivery
system for articles to scientists in developing countries. The list of eligible
countries (most, if not all, African countries are included) and information on
how to subscribe can be found at http://library.ictp.it/ejds
Downloads are limited to 3 papers per day, 12 per month and 100 per year.
:: HIGH PERFORMANCE COMPUTING SCHOOL ::
CHPC and IBM South Africa are pleased to announce they will be sponsoring a
limited number of students from outside South Africa to attend the High
Performance Computing School and CHPC National Conference 2010 from the 29
November 2010 - 10 December 2010. The course will cover the concepts and
theory of parallel computers, and programming for parallel systems with MPI,
OpenMP and CUDA, using the C, Fortran or python programming languages.
Please see http://www.chpcconf.co.za/index.cfm?x=hpcs for further details.
Deadline for applications: 29 October 2010
:: RESEARCH ARTICLE OF THE MONTH ::
Last month's Psi-k Highlight of the Month 'Ab Initio Random Structure
Searching' by Chris J Pickard and R J Needs
http://www.psi-k.org/newsletters/News_100/Highlight_100.pdf
Predicting the configuration that a collection of atoms will adopt is
one of the most common problems for computational scientists. Usually,
as in the case of the silica project, we begin with knowledge of the
structure from direct experimental observation. However in the cases
where experimental evidence is inconclusive or simply does not exist
(which can be the case for extreme circumstances such as very high
pressure or temperature), we don't have a structure to begin with!
Further to this, we could also be interested in metastable structures
that could be obtained by non-equilibrium conditions. A common approach to
this problem is to use simulated annealing where the system is repeatedly
heated and cooled to allow the atoms to find their lowest energy
configuration. Last month's Psi-K highlight of the month explained how
Chris Pickard and Richard Needs developed an alternative approached called
'Ab Initio Random Structure Searching' (AIRSS). It involves creating 'random'
atomic configurations for the material and calculating which arrangements
yield energy minima.
In order to find the lowest energy structures, we need to identify the local
minima of a potential energy surface. The number of local minima has been
shown to increase exponentially with the number of atoms. The task to calculate
the energy of each of the possible local minima essentially becomes impossible.
The idea behind AIRSS is starting with a 'sensible' random structure, where some
initial constraints are opposed on the beginning structure that will hopefully
not bias the calculation to the point where it does not explore enough of the
potential energy surface. General assumptions that are made for all systems are
that they are always fully connected, and so each atom is always connected to at
least one other atom. In addition to this, physically or chemically unreasonable
configurations are discounted, such as having bond lengths that are far too
small. The paper goes on to explain the various constraints that are employed in
studying crystals, cluster and defects.
Two of the big successes of the AIRSS method have involved the search for
metallic hydrogen. Its first big achievement was the prediction of previously
unknown phases of pure hydrogen at high pressures. Previous DFT calculations had
predicted that above about 200GPa, a transition to a metallic phase would occur.
This was contrary to experimental data which showed that it remained insulating
well above 300 GPa. AIRSS calculations showed up several alternative structures
that were in fact insulating. Furthering the search for metallic hydrogen,
calculations on possible high energy phases of aluminium hydride were carried out.
Pickard and Needs predicted a stable semimetallic phase which was later confirmed
by high pressure diffraction experiments. The paper describes the application of
AIRSS to a plethora of other systems, including more elemental and compound
structures at high pressures, the identification of metastable structures, and the
calculation of defects in silicon, showing the extremely wide applicability of the
method.
:: GROUP PROJECT UPDATE ::
We now have the data collected from what was calculated in Cape Town!
Richard Martin sent out a full summary of what was done, and so if you
haven't received it please let him know. Below is a brief outline of Stefano
di Gironcoli's collected results from the school.
Calculations of the static equations of state for silica (SiO2) in the Quartz
and Stishovite structure using five different exchange-correlation functionals
were carried out. Two of these were LDA (PZ and VWN) and three were GGA (PBE,
BLYP and WC). Everyone was assigned a functional to calculate and the two
structures were optimized for a pressure range of P=-30 kBar to P=200 kBar. The
results generally agreed with each other with the following results for the
transition pressures:
PZ: 30-35 kBar
VWN: 30-35 kBar
PBE: 95-100 kBAR
WC: 55-60 kBar
BLYP: 155 kBar
These gave the correct ordering of the transition pressures for the various
functionals, however with a difference of ~30 kBar from published results. The
reasons for this offset are now being examined.
The second part of the project involved calculating the dynamical matrices and
phonon frequencies of the quartz and stishovite structures. Enthalpy-Pressure
diagrams were calculated for static, T=0K, T=300K and T=500K. Unfortunately, not
all results for this part were completed due to the more computationally demanding
nature of the calculations.
However, the current results give the transition pressures below:
Temp PZ VWN PBE WC
-------------------------------
Static 26.2 32.0 98.6 56.4
0K 27.0 32.9 98.6 58.2
300K 31.1 36.9 103.2 61.8
500K 37.7 42.2 108.0 66.4
We are now continuing with the project so those that would like to be
involved but haven't contacted Richard should do so as soon as possible.