2.3.2.3 Isotropic Distribution Function

Eqns. (2.76) to (2.78) all contain a statistical average of a symmetric tensor (see Appendix A) of the form $ \ensuremath{\langle \ensuremath{\boldsymbol{\mathrm{p}}} \ensuremath{\otimes}\ensuremath{\boldsymbol{\mathrm{p}}} \rangle}$ which will be evaluated in this section. The distribution function can be decomposed into a symmetric and an anti-symmetric part

$\displaystyle f(\ensuremath{\boldsymbol{\mathrm{k}}}) = f_\mathrm{S}(\ensuremat...
...ldsymbol{\mathrm{k}}}) + f_\mathrm{A}(\ensuremath{\boldsymbol{\mathrm{k}}}) \ .$ (2.79)

In this section we assume that the symmetric part is isotropic

$\displaystyle f(\ensuremath{\boldsymbol{\mathrm{k}}}) = f_\mathrm{S}(\vert\ensu...
...bol{\mathrm{k}}}\vert) + f_\mathrm{A}(\ensuremath{\boldsymbol{\mathrm{k}}}) \ .$ (2.80)

This is a special case of the diffusion approximation [21, p.49] which will be explained in more detail for the MAXWELL distribution in Section 2.3.3.1 on page [*]. By using this assumption the statistical average of the tensor can be written as

$\displaystyle \ensuremath{\langle \ensuremath{\boldsymbol{\mathrm{p}}} \ensurem...
...\ensuremath{\boldsymbol{\mathrm{k}}}\vert) \,\, \ensuremath{\mathrm{d}}^3 k}\ .$ (2.81)

For symmetry reasons all elements outside the trace vanish. For instance, the element

$\displaystyle \ensuremath{\langle k_x \, k_y \rangle}= \ensuremath{\iiint\limit...
...ath{\mathrm{d}}k_x \, \ensuremath{\mathrm{d}}k_y \, \ensuremath{\mathrm{d}}k_z}$ (2.82)

evaluates to zero because of the integral

$\displaystyle \int_{-\infty}^{\infty} k_x \, f_\mathrm{S} \left( \sqrt{k_x^2 + k_y^2 + k_z^2} \right) \,\, \mathrm{d} k_x = 0 \ .$ (2.83)

Since the distribution function is assumed to be isotropic, the integrals determining the elements of the trace all evaluate to a common value $ J$

$\displaystyle \ensuremath{\langle k_l \, k_l \rangle}= \ensuremath{\iiint\limit...
...ensuremath{\mathrm{d}}k_z}= J \ , \textcolor{lightgrey}{.......}l = x, y, z \ .$ (2.84)

The value of $ J$ can be evaluated by the simple transformation

$\displaystyle \ensuremath{\langle k_x \, k_x \rangle}= \ensuremath{\langle k_y \, k_y \rangle}= \ensuremath{\langle k_z \, k_z \rangle}$ $\displaystyle = J \ ,$ (2.85)
$\displaystyle \ensuremath{\langle k_x^2 \rangle}+ \ensuremath{\langle k_y^2 \ra...
...math{\langle k_z^2 \rangle}= \ensuremath{\langle k_x^2 + k_y^2 + k_z^2 \rangle}$ $\displaystyle = 3 \, J \ ,$ (2.86)
$\displaystyle \frac{\ensuremath{\langle k^2 \rangle}}{3}$ $\displaystyle = J \ .$ (2.87)

Therefore, the statistical averages of the tensors are diagonal with all diagonal elements being equal:

$\displaystyle \ensuremath{\langle \ensuremath{\boldsymbol{\mathrm{p}}} \ensuremath{\otimes}\ensuremath{\boldsymbol{\mathrm{p}}} \rangle}$ $\displaystyle = \frac{{\ensuremath{\langle p^2 \rangle}}}{3} \, \ensuremath{\widetilde{\delta}}\ ,$ (2.88)
$\displaystyle \ensuremath{\langle \ensuremath{\boldsymbol{\mathrm{p}}} \ensuremath{\otimes}\ensuremath{\boldsymbol{\mathrm{p}}} \, \mathcal{E} \rangle}$ $\displaystyle = \frac{{\ensuremath{\langle \mathcal{E}\, p^2 \rangle}}}{3} \, \ensuremath{\widetilde{\delta}}\ .$ (2.89)

By inserting eqns. (2.88) and (2.89) into eqns. (2.76) to (2.78) one gets

  $\displaystyle \ensuremath{\boldsymbol{\mathrm{\phi}}}_1:$ $\displaystyle \textcolor{lightgrey}{.......}$ $\displaystyle \frac{2}{3} \, \ensuremath{\ensuremath{\ensuremath{\boldsymbol{\mathrm{\nabla}}}}\, \ensuremath{\langle \mathcal{E} \rangle}}$   $\displaystyle + q \, \ensuremath{\boldsymbol{\mathrm{E}}}\, \ensuremath{\langle 1 \rangle}$   $\displaystyle = - m \, \frac{\ensuremath{\langle \ensuremath{\boldsymbol{\mathrm{v}}} \rangle}}{\tau_m} \ ,$ (2.90)
  $\displaystyle \ensuremath{\boldsymbol{\mathrm{\phi}}}_3:$ $\displaystyle \textcolor{lightgrey}{.......}$ $\displaystyle \frac{1}{3} \, \ensuremath{\ensuremath{\ensuremath{\boldsymbol{\mathrm{\nabla}}}}\, \ensuremath{\langle v^2 \, \mathcal{E} \rangle}}$   $\displaystyle + \frac{5}{3} \, \frac{\mathrm{q}}{m} \, \ensuremath{\boldsymbol{\mathrm{E}}}\, \ensuremath{\langle \mathcal{E} \rangle}$   $\displaystyle = - \frac{\ensuremath{\langle \ensuremath{\boldsymbol{\mathrm{v}}} \, \mathcal{E} \rangle}}{\tau_S} \ ,$ (2.91)
  $\displaystyle \ensuremath{\boldsymbol{\mathrm{\phi}}}_5:$ $\displaystyle \textcolor{lightgrey}{.......}$ $\displaystyle \frac{1}{3} \, \ensuremath{\ensuremath{\ensuremath{\boldsymbol{\mathrm{\nabla}}}}\, \ensuremath{\langle v^2 \, v^2 \, \mathcal{E} \rangle}}$   $\displaystyle + \frac{7}{3} \, \frac{\mathrm{q}}{m} \, \ensuremath{\boldsymbol{\mathrm{E}}}\, \ensuremath{\langle v^2 \, \mathcal{E} \rangle}$   $\displaystyle = - \frac{\ensuremath{\langle \ensuremath{\boldsymbol{\mathrm{v}}} \, v^2 \, \mathcal{E} \rangle}}{\tau_K} \ .$ (2.92)

Note that the divergences of the tensors simplify to gradients of scalars.

M. Gritsch: Numerical Modeling of Silicon-on-Insulator MOSFETs PDF