24.10.2014 Views

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

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

CHAPTER 9.<br />

ENERGY<br />

A NEW CALCULATION OF THE JELLIUM SURFACE<br />

9.3 Surface energy calculations<br />

On reviewing the earlier sections of this <strong>thesis</strong>, it is apparent that even with the<br />

many improvements introduced here, finite-size errors cannot be eliminated. In<br />

particular, the Coulomb finite-size error conjectured to be caused by ‘squashing’ of<br />

the exchange-correlation hole must still be dealt with. In order to minimise the effect<br />

of these errors on the calculation of the surface energy, great care must be taken<br />

with the extrapolation to infinite system size; in addition, it is unwise to compare<br />

results from slab and bulk systems.<br />

This latter constraint means that simulations must be carried out for a range<br />

of slab widths; the bulk and surface energies may then be extracted by a fitting<br />

procedure, using equation (5.6), repeated here for convenience:<br />

( ) 8πr<br />

3<br />

ɛ slab = ɛ bulk + s σ 1<br />

3 s . (9.6)<br />

The motivation for this approach is that the errors arising in bulk and slab calculations<br />

may differ significantly; using only slab results should ensure a better<br />

cancellation of errors.<br />

As was noted in chapter 5, the surface energy calculated using equation (9.6)<br />

displays oscillations of decreasing amplitude as the slab width is increased; the true<br />

surface energy is the limiting value of σ as s → ∞. To minimise the errors introduced<br />

by these oscillations, it is possible to choose slab widths for which the LDA value<br />

of σ matches the extrapolated value: three such special slab widths are 11.7783,<br />

15.1317 and 18.4851.<br />

For each slab width, calculations must be performed for a range of cell sizes; the<br />

results can then be extrapolated to the limit of infinite cell size. By selecting sizes<br />

for which the LDA slab energy per electron corresponds to the infinite-cell value for<br />

the same slab width, the need to apply the independent-particle finite-size correction<br />

described in section 4.1.1 is obviated.<br />

Each system to be simulated requires an optimised trial wave function. As usual,<br />

LDA orbitals (with the image-tail correction of section 9.1) make up the Slater<br />

174

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!