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 4.<br />
ERRORS IN QMC SIMULATIONS<br />
There is an infinite number of permissible k-vectors in the first Brillouin zone;<br />
imposing strictly periodic boundary conditions on the wave function, rather than<br />
on Ĥ, corresponds to taking k = 0.<br />
However, there is no a priori reason why this particular k-point should be better<br />
than any other. In a non-interacting system, the single-electron wave functions are<br />
plane waves, and the infinite-system limit is obtained by integrating over all k-<br />
points. In an interacting system, this may also be the ideal approach [55]. The aim<br />
is always to choose the k-point(s) which best reproduce(s) the infinite-system result;<br />
several studies have investigated ways to achieve this goal in the context of QMC<br />
simulations [74, 39]. The use of alternative k-points is also discussed in chapter 9.<br />
Another way to correct this kind of finite-size error in QMC results is to apply a<br />
correction of the form (E DFT<br />
∞<br />
− E DFT<br />
N<br />
), where EDFT<br />
N<br />
and E DFT<br />
∞<br />
are the DFT results<br />
for finite and infinite cells respectively. This is a valid procedure because finite-cell<br />
DFT calculations suffer from the same k-point sampling errors, and it is usually<br />
easier to extrapolate the results of DFT calculations to the infinite-cell limit, thus<br />
obtaining E∞<br />
DFT . However, DFT calculations do not suffer from the second type of<br />
finite-size error: that related to the Coulomb interaction.<br />
4.1.2 Coulomb finite-size errors<br />
A requirement of any QMC simulation is the ability to determine the Coulomb<br />
energy of a given configuration. In a finite system, this is trivial; in a finite system<br />
which is designed to model an infinite system, it is not.<br />
Referring to figure 4.1, each ‘real’ electron in the simulation cell must interact<br />
with all the others; there must also be some interaction between these electrons and<br />
the ‘imaginary’ electrons outside the simulation cell.<br />
The conventional approach to this problem is to solve Poisson’s equation for the<br />
charges in the cell with periodic boundary conditions. 4 The result is known as the<br />
4 Note that the simulation cell as a whole must have zero net charge; all the electron charges<br />
must be cancelled by an equivalent amount of positive charge (provided by ions in real materials,<br />
61