04.08.2013 Views

Statistical Mechanics - Physics at Oregon State University

Statistical Mechanics - Physics at Oregon State University

Statistical Mechanics - Physics at Oregon State University

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.

5.1. FERMIONS IN A BOX. 91<br />

Znx,ny,nz(T, µ, V ) = 1 + e µ−ɛ(nx,ny ,nz )<br />

k B T (5.8)<br />

We now introduce again λ = e µ<br />

k B T , which is always positive, and if we are<br />

allowed to replace the sum by an integral we obtain:<br />

1<br />

V Ω(T, µ, V ) = −(2S + 1)(2π)−3 <br />

kBT<br />

d 3 k log(1 + λe − 2 k 2<br />

2Mk B T ) (5.9)<br />

Formulas like these make it clear th<strong>at</strong> we really want to use kB for the Boltzmann<br />

constant, in order to avoid confusion with the norm of the wave vector, k.<br />

Free particles have a simple volume dependence!<br />

One remarkable thing happened. The right hand side does not depend on<br />

volume anymore! When we had a summ<strong>at</strong>ion over discrete k-vectors, the volume<br />

dependence was included in these vectors. But when we changed to a continuous<br />

summ<strong>at</strong>ion (or integr<strong>at</strong>ion), this volume dependence disappeared, and the only<br />

volume dependence is a simple pre-factor V , a linear behavior!<br />

Those m<strong>at</strong>hem<strong>at</strong>ical details!<br />

Wh<strong>at</strong> could go wrong here? There is an error associ<strong>at</strong>ed with the change<br />

from summ<strong>at</strong>ion to integr<strong>at</strong>ion, and in general we have:<br />

<br />

1<br />

Ω(T, µ, V ) = −(2S + 1)(2π)<br />

V −3 <br />

kBT<br />

d 3 k log(1 + λe − 2k 2<br />

<br />

2MkB T ) [1 + E(T, V, µ)]<br />

(5.10)<br />

where we know th<strong>at</strong> limV →∞ E(T, V, µ) = 0. But we would like to be able to<br />

differenti<strong>at</strong>e the result with respect to T and µ, and hence we need to require<br />

uniform convergence! But th<strong>at</strong> in turn also begs the question if the original<br />

series is uniformly convergent! In order to discuss the convergence properties<br />

we have to take two steps. First, we have to show th<strong>at</strong> the series converges<br />

uniformly. Second, we have to show th<strong>at</strong> the series in the thermodynamic limit<br />

converges uniformly to the integral. We will discuss the uniform convergence of<br />

the series first. The important thing to notice is th<strong>at</strong> both the integral and the<br />

series converge.<br />

For large values of the wave vector the terms in the series behave like:<br />

tnx,ny,nz = log(1 + λe − ɛ(nx,ny ,nz )<br />

k B T ) ≈ λe − ɛ(nx,ny ,nz )<br />

k B T (5.11)<br />

which approaches zero exponentially fast. Next we consider a domain in (T, µ, L)<br />

space with 0 T Tm , −∞ µ µm , and 0 L Lm. It is clear th<strong>at</strong> for<br />

large values of the indices we have

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

Saved successfully!

Ooh no, something went wrong!