My PhD thesis - Condensed Matter Theory - Imperial College London
My PhD thesis - Condensed Matter Theory - Imperial College London
My PhD thesis - Condensed Matter Theory - Imperial College London
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
CHAPTER 5.<br />
THE JELLIUM SLAB<br />
reducing the average kinetic energy at the expense of increasing the potential energy<br />
by a larger amount.<br />
The problem occurs both when reweighting of configurations is used and when<br />
it is not. This suggests that the cause is not outliers (configurations with energy far<br />
from the mean). Figure 5.7 shows that the energy distributions are close to being<br />
Gaussian, with only slightly ‘fat’ upper tails.<br />
The configurations sampled initially do not include any with electrons far outside<br />
the slab; the electron density decays sharply to zero here, as can be seen in figure 5.1.<br />
After the first stage of variance minimisation, the Jastrow factor is altered so that<br />
the electron density extends much further outside the slab; therefore the next set<br />
of configurations which are generated will sample a different region of configuration<br />
space to the old set. The effect of pushing many electrons into the vacuum region,<br />
and the consequential increase in the potential energy, cannot be ‘known’ in advance<br />
by the variance minimisation routine: no such configurations have been sampled.<br />
This suggests that the process may be corrected by modifying the initial sampling<br />
distribution to include these configurations. Unfortunately, this approach did not<br />
work; the procedure remains unstable, with the mean and variance of the local<br />
energy usually worsening after each iteration.<br />
Without reweighting, the object of the variance minimisation step is to minimise<br />
the following quantity:<br />
O({X i ; α}) = ∑ i<br />
[<br />
E L (X i ; α) − 1 N<br />
∑<br />
2<br />
E L (X j ; α)]<br />
. (5.8)<br />
j<br />
As before, α denotes the optimiseable parameter, {X i } is the set of configurations<br />
and E L is the local energy. This may be split into kinetic and potential terms; the<br />
potential energy does not depend on α:<br />
O({X i ; α}) = ∑ i<br />
[<br />
T i (α) − ¯T (α) + V i − ¯V<br />
] 2<br />
. (5.9)<br />
82