- #1
Philethan
- 35
- 4
I'm reading Mermin's Solid State Physics, chapter 12: The semiclasssical model of electron dynamics. I know the current density from the ##n## band is
$$ \mathbf{j}=(-e)\int_{\text{occupied}}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}\mathbf{v}_{\text{n}}(\mathbf{k}). $$
In addition, for the velocity component which is perpendicular to magnetic field, it is:
$$ \mathbf{v}_{\text{n},\perp } \approx\frac{\mathbf{r}_{\perp}(0)-\mathbf{r}_{\perp}(-\tau)}{\tau}=-\frac{\hbar c}{eH}\hat{H}\times\frac{\mathbf{k}(0)-\mathbf{k}(-\tau)}{\tau}+c\frac{E}{H}(\hat{E}\times\hat{H})$$
Therefore,
$$\mathbf{j}_{\perp}\approx(-e)\int_{\text{occupied}}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}\underbrace{\left\{\left[-\frac{\hbar c}{eH}\hat{H}\times\frac{\mathbf{k}(0)-\mathbf{k}(-\tau)}{\tau}+c\frac{E}{H}(\hat{E}\times\hat{H})\right]\right\}}_{\mathbf{v}_{\text{n},\perp}(\mathbf{k})}$$
If some of the occupied and unoccupied levels lie on orbits that do not close on themselves, but are extended or "open" in ##k##-space, then the first term at the right-hand side of ##\mathbf{v}_{\text{n},\perp}## cannot be neglected. Therefore, the following derivation would be incorrect:
$$\mathbf{j}_{\perp}\approx(-e)(\mathbf{v}_{\text{n},\perp})\underbrace{\left[\int_{\text{occupied}}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}\right]}_{n(\text{electron carrier number density})}=-ne\mathbf{v}_{\text{n},\perp}$$
My first question is, I think in the above case, ##\mathbf{j}_{\perp}\neq-ne\mathbf{v}##, and thus ##\mathbf{j}_{\perp}\neq \mathbf{\sigma}\cdot\mathbf{E}##. Is that correct? I know the reason why ##\mathbf{j}_{\text{n}}=\mathbf{\sigma}\cdot\mathbf{E}## holds is the semiclassical equations of motion:
$$\mathbf{v}_{\text{n}}(\mathbf{k})=\frac{1}{\hbar}\nabla_{\mathbf{k}}\epsilon_{\text{n}}(\mathbf{k})$$
and
$$\hbar\dot{\mathbf{k}}=(-e)\left[\mathbf{E}(\mathbf{r},t)+\frac{1}{c}\mathbf{v}_{\text{n}}(\mathbf{k})\times\mathbf{H}(\mathbf{r},t)\right]$$
Therefore, perhaps we can approximate ##\mathbf{v}_{n}(\mathbf{k})## as this:
$$\mathbf{v}_{n}(\mathbf{k})\approx\cfrac{\int_{occupied}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}\mathbf{v}_{\text{n}}(\mathbf{k})}{\int_{\text{occupied}}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}}=\cfrac{\int_{occupied}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}\mathbf{v}_{\text{n}}(\mathbf{k})}{n_{\text{electron}}}$$However, how to know its error? How to judge this approximation?And my second question is, from (12.56)
$$\mathbf{j}=\sigma^{(0)}\hat{\mathbf{n}}(\hat{\mathbf{n}}\cdot\mathbf{E})+\mathbf{\sigma}^{(1)}\cdot\mathbf{E},\begin{cases}\sigma^{(0)}\to\text{constant as }H\to\infty\\\sigma^{(1)}\to0\text{ as }H\to\infty\end{cases}$$
It seems that the current flow "always" lie along the direction ##\hat{\mathbf{n}}## of the open orbit in real space (I think ##\hat{n}## is the direction of ##\hat{H}\times\left[\mathbf{k}(0)-\mathbf{k}(-\tau)\right]##), but authors said
It makes me think ##\mathbf{E}\cdot\hat{\mathbf{n}}## always naturally vanish. Does authors talk about a special case? Or ##\mathbf{E}\cdot\hat{\mathbf{n}}## really "always" vanish? If it's just a special case, then how to know whether ##\mathbf{E}\cdot\hat{\mathbf{n}}## is zero?Sorry for my bad English, thank you so much!
$$ \mathbf{j}=(-e)\int_{\text{occupied}}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}\mathbf{v}_{\text{n}}(\mathbf{k}). $$
In addition, for the velocity component which is perpendicular to magnetic field, it is:
$$ \mathbf{v}_{\text{n},\perp } \approx\frac{\mathbf{r}_{\perp}(0)-\mathbf{r}_{\perp}(-\tau)}{\tau}=-\frac{\hbar c}{eH}\hat{H}\times\frac{\mathbf{k}(0)-\mathbf{k}(-\tau)}{\tau}+c\frac{E}{H}(\hat{E}\times\hat{H})$$
Therefore,
$$\mathbf{j}_{\perp}\approx(-e)\int_{\text{occupied}}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}\underbrace{\left\{\left[-\frac{\hbar c}{eH}\hat{H}\times\frac{\mathbf{k}(0)-\mathbf{k}(-\tau)}{\tau}+c\frac{E}{H}(\hat{E}\times\hat{H})\right]\right\}}_{\mathbf{v}_{\text{n},\perp}(\mathbf{k})}$$
If some of the occupied and unoccupied levels lie on orbits that do not close on themselves, but are extended or "open" in ##k##-space, then the first term at the right-hand side of ##\mathbf{v}_{\text{n},\perp}## cannot be neglected. Therefore, the following derivation would be incorrect:
$$\mathbf{j}_{\perp}\approx(-e)(\mathbf{v}_{\text{n},\perp})\underbrace{\left[\int_{\text{occupied}}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}\right]}_{n(\text{electron carrier number density})}=-ne\mathbf{v}_{\text{n},\perp}$$
My first question is, I think in the above case, ##\mathbf{j}_{\perp}\neq-ne\mathbf{v}##, and thus ##\mathbf{j}_{\perp}\neq \mathbf{\sigma}\cdot\mathbf{E}##. Is that correct? I know the reason why ##\mathbf{j}_{\text{n}}=\mathbf{\sigma}\cdot\mathbf{E}## holds is the semiclassical equations of motion:
$$\mathbf{v}_{\text{n}}(\mathbf{k})=\frac{1}{\hbar}\nabla_{\mathbf{k}}\epsilon_{\text{n}}(\mathbf{k})$$
and
$$\hbar\dot{\mathbf{k}}=(-e)\left[\mathbf{E}(\mathbf{r},t)+\frac{1}{c}\mathbf{v}_{\text{n}}(\mathbf{k})\times\mathbf{H}(\mathbf{r},t)\right]$$
Therefore, perhaps we can approximate ##\mathbf{v}_{n}(\mathbf{k})## as this:
$$\mathbf{v}_{n}(\mathbf{k})\approx\cfrac{\int_{occupied}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}\mathbf{v}_{\text{n}}(\mathbf{k})}{\int_{\text{occupied}}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}}=\cfrac{\int_{occupied}f(\epsilon_{\text{n}}(\mathbf{k}))\frac{d\mathbf{k}}{4\pi^3}\mathbf{v}_{\text{n}}(\mathbf{k})}{n_{\text{electron}}}$$However, how to know its error? How to judge this approximation?And my second question is, from (12.56)
$$\mathbf{j}=\sigma^{(0)}\hat{\mathbf{n}}(\hat{\mathbf{n}}\cdot\mathbf{E})+\mathbf{\sigma}^{(1)}\cdot\mathbf{E},\begin{cases}\sigma^{(0)}\to\text{constant as }H\to\infty\\\sigma^{(1)}\to0\text{ as }H\to\infty\end{cases}$$
It seems that the current flow "always" lie along the direction ##\hat{\mathbf{n}}## of the open orbit in real space (I think ##\hat{n}## is the direction of ##\hat{H}\times\left[\mathbf{k}(0)-\mathbf{k}(-\tau)\right]##), but authors said
Because of (12.56) this is possible in the high-field limit only if the projection of the electric field on ##\hat{\mathbf{n}}##, ##\mathbf{E}\cdot\hat{\mathbf{n}}##, vanishes. (page 237-238)
It makes me think ##\mathbf{E}\cdot\hat{\mathbf{n}}## always naturally vanish. Does authors talk about a special case? Or ##\mathbf{E}\cdot\hat{\mathbf{n}}## really "always" vanish? If it's just a special case, then how to know whether ##\mathbf{E}\cdot\hat{\mathbf{n}}## is zero?Sorry for my bad English, thank you so much!