14.09.2014 Views

CASINO manual - Theory of Condensed Matter

CASINO manual - Theory of Condensed Matter

CASINO manual - Theory of Condensed Matter

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

33.2.2 Atomic densities<br />

K eywords: density or spin density (both produce the same thing)<br />

Density accumulation for finite systems is only available for single atoms; the molecular case has not<br />

been coded in casino.<br />

In the single-atom case, the charge density is obviously spherically symmetric, ρ(r) = ρ(r).<br />

spherical average <strong>of</strong> the charge density is<br />

ρ sph (r) = 1<br />

4πr 2<br />

N ∑<br />

i=1<br />

The<br />

∫<br />

f(R)δ(|ri | − r)dR<br />

∫<br />

f(R)dR<br />

, (307)<br />

where ∫ f(R) is the distribution <strong>of</strong> configurations.<br />

4πr 2 ρ sph (r)dr = N.<br />

The charge density is normalized such that<br />

The analytical behaviour <strong>of</strong> ρ(r), ρ sph (r) at r → 0 and r → ∞ is known,<br />

ρ ′ sph(0) = −2Zρ sph (0) , (308)<br />

where Z is the atomic number, and<br />

(<br />

ρ sph (r) → Ar 2B exp −2 √ )<br />

2Ir , (309)<br />

r→∞<br />

where A is a constant, B = (Z − N + 1)/ √ 2I − 1, and I is the ionization energy.<br />

The method used to calculate the charge density in QMC is presented below for the general case <strong>of</strong><br />

the charge density, but it is easily applied to quantities such as the spherical average <strong>of</strong> the charge<br />

density.<br />

Suppose we are performing a QMC calculation, sampling a distribution f(R). In the case <strong>of</strong> VMC,<br />

f(R) = |Ψ(R)| 2 , while in DMC the distribution is f(R) = Ψ(R)Φ(R). The charge density can be<br />

evaluated in QMC as<br />

∑ M<br />

m=1<br />

ρ(r) ≈ ¯ρ(r) =<br />

w mρ(R m )<br />

∑ M<br />

m=1 w , (310)<br />

m<br />

where {R m } M m=1 are the M configurations visited during the simulation and w m is the weight <strong>of</strong> each<br />

<strong>of</strong> them, which is constant in VMC and varies from configuration to configuration in DMC. The<br />

‘approximately equal’ sign becomes an ‘equal’ sign in the limit <strong>of</strong> perfect sampling, that is, when<br />

M → ∞.<br />

We can identify ∑ i δ(r i − r) with ρ(R), resulting in<br />

∑ M ∑<br />

m=1 i<br />

ρ(r) ≈ ¯ρ(r) = w mδ(r i,m − r)<br />

∑ M<br />

m=1 w . (311)<br />

m<br />

However, Eq. (311) is not useful for accumulating ¯ρ(r) since Dirac delta functions cannot be evaluated<br />

outside an integral.<br />

We overcome this by accumulating data in bins <strong>of</strong> finite width. Let us partition three-dimensional<br />

space into N exp disjoint regions or bins {Ω p }, and let us define the step function<br />

{ 1 if r ∈ Ωp<br />

Θ p (r) =<br />

. (312)<br />

0 if r /∈ Ω p<br />

Denoting the volume <strong>of</strong> region Ω p by the same symbol, Ω p , we can construct an orthogonal functional<br />

basis h p (r) = Ω −1/2<br />

p Θ p (r). Then, Eq. (311) becomes<br />

¯ρ(r) ≈<br />

N exp<br />

∑<br />

p=1<br />

that is, the function inside Ω p is approximated by a single value<br />

∑ M ∑<br />

m=1 i w mΘ p (r i,m )<br />

∑ M<br />

Ω p m=1 w Θ p (r) , (313)<br />

m<br />

∑ M ∑<br />

m=1 i<br />

ρ p = w mΘ p (r i,m )<br />

∑ M<br />

Ω p m=1 w . (314)<br />

m<br />

185

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

Saved successfully!

Ooh no, something went wrong!