2.5.5.3 Screening Theory

The screening effect is the most important manifestation of the electron-electron interaction. In this work only linear screening is considered. Nonlinear effects which require the solution of the nonlinear Poisson equation either analytically [46,47] or numerically [48] are neglected here.

A positively charged particle introduced into the electron gas will create an excess of negative charge in its vicinity which screen its electric field2.35. The bare potential (2.142) satisfies the Poisson equation:

$\displaystyle \varepsilon\nabla^{2}V_{0}(\vec{r})=-\rho_{0}(\vec{r}),$ (2.144)

where $ \rho_{0}(\vec{r})=Ze[\delta(\vec{r})+\delta(\vec{r}-\vec{R})]$ is the volume density of the particle charge2.36. The total physical potential $ V_{t}(\vec{r})$, which is produced both by the particle and the electron cloud, is described by the Poisson equation of the form:

$\displaystyle \varepsilon\nabla^{2}V_{t}(\vec{r})=-\rho(\vec{r}),$ (2.145)

where $ \rho(\vec{r})$ is the total volume charge density:

$\displaystyle \rho(\vec{r})=\rho_{0}(\vec{r})+\rho_\mathrm{ind}(\vec{r}).$ (2.146)

Here $ \rho_\mathrm{ind}(\vec{r})$ denotes the volume density of the charge induced in the electron gas by the positive particle. From the electrodynamics of continuous media [49] it is known that $ V_{0}(\vec{r})$ and $ V_{t}(\vec{r})$ are linearly related to each other:

$\displaystyle V_{0}(\vec{r})=\int \varepsilon(\vec{r},\vec{r}^{'})V_{t}(\vec{r}^{'})\,d\vec{r}^{'}.$ (2.147)

For the homogeneous electron gas function $ \varepsilon(\vec{r},\vec{r}^{'})$ can only depend on the distance between the two points $ \vec{r}$ and $ \vec{r}^{'}$:

$\displaystyle \varepsilon(\vec{r},\vec{r}^{'})=\varepsilon(\vec{r}-\vec{r}^{'}).$ (2.148)

This gives the following relation between the Fourier components of $ V_{0}(\vec{r})$ and $ V_{t}(\vec{r})$:

$\displaystyle V_{t}(\vec{q})=\frac{V_{0}(\vec{q})}{\varepsilon(\vec{q})}.$ (2.149)

In the linear screening theory it is assumed that $ \rho_\mathrm{ind}(\vec{r})$ and $ V_{t}(\vec{r})$ are linearly related to each other2.37 which implies for Fourier components:

$\displaystyle \rho_\mathrm{ind}(\vec{q})=\chi(\vec{q})V_{t}(\vec{q}).$ (2.150)

The Fourier transform of the Poisson equations (2.144) and (2.145) together with (2.150) leads to the relation:

$\displaystyle \varepsilon(\vec{q})=1-\frac{1}{\vec{q}^{2}}\chi(\vec{q}).$ (2.151)

To calculate the quantity $ \chi(\vec{q})$ some approximations are necessary. In this work we use the so called Lindhard screening theory2.38. Within this model the expression for the dielectric function [50] is:

$\displaystyle \varepsilon(\vec{q},0)=1+\frac{\beta^{2}_{s}}{q^{2}}G\left(\xi,\eta\right),$ (2.152)

with the inverse screening length given by the Thomas-Fermi theory:

$\displaystyle \beta^{2}_{s}=\frac{e^{2}n}{\varepsilon k_{B}T_{L}} \frac{\mathcal{F}_{-1/2}\left(\eta\right)}{\mathcal{F}_{1/2}\left(\eta\right)}.$ (2.153)

Here $ \mathcal{F}_{i}$ stands for the Fermi integral of the $ i$-th order, and $ \eta$ is the reduced Fermi energy:

$\displaystyle \eta=\frac{E_{F}-E_{C}}{k_{B}T_{L}}.$ (2.154)

$ G$ represents the screening function given by the following expression:

$\displaystyle G\left(\xi,\eta\right)=\frac{1}{\mathcal{F}_{-1/2}\left(\eta\righ...
...1+\exp\left(x^{2}-\eta\right)}\ln\biggl\vert\frac{x+\xi}{x-\xi}\biggr\vert\,dx,$ (2.155)

where the argument $ \xi$ is:

$\displaystyle \xi^{2}=\frac{\hbar^{2}q^{2}}{8m^{*}k_{B}T_{L}}.$ (2.156)

The corrections described above can be taken into account within the first Born approximation. In this approximation the scattering amplitude is given as:

$\displaystyle f\left(\vec{q}\right)=-\frac{1}{4\pi}U\left(\vec{q}\right),$ (2.157)

where $ U(\vec{q})$ is defined by

$\displaystyle U(\vec{q})=-\frac{2eV_{t}(\vec{q})m^{*}}{\hbar^{2}}.$ (2.158)

The differential cross section is defined as:

$\displaystyle \sigma(\vec{q})=\frac{(2\pi\hbar)^{3}}{m^{*2}\vert\vec{v}(\vec{k})\vert}\vert f(\vec{q})\vert^{2}\rho(\epsilon).$ (2.159)

Now the total scattering rate is obtained through

$\displaystyle \lambda(\vec{k})=N_{p}\vert\vec{v}(\vec{k})\vert\sigma_\mathrm{tot}(\vec{k}),$ (2.160)

where $ N_{p}=N_{I}/2$ is the density of scattering pairs and $ \sigma_\mathrm{tot}(\vec{k})$ is equal to:

$\displaystyle \sigma_\mathrm{tot}(\vec{k})=\frac{2\pi}{\vec{k}^{2}}\int\sigma(\vec{q})q\,dq.$ (2.161)

The simple calculations give the final total ionized impurity scattering rate taking into account the momentum dependent screening and the two-ion correction:
    $\displaystyle \lambda(\vec{k})=C\left(k\right)
\int_{0}^{2k}\frac{1}{(q^{2}+\be...
...}G\left(\xi,\eta\right))^{2}}\left(1+\frac{\sin\left(qR\right)}{qR}\right)\,dq,$  
    $\displaystyle C\left(k\right)=\frac{N_{i}Z^{2}e^{4}}{2\pi\hbar^{2}\varepsilon^{2}\vert\vec{v}(\vec{k})\vert}.$ (2.162)

S. Smirnov: