Hi I am going through Fetter's Quantum Theory of Many-Particle Systems Dover Edition.

In ch. 31 he computed the relation between $\bar{G}(\mathbf{k},\omega)$, ${\bar{G}}^{R}(\mathbf{k},\omega)$ and $\bar{G}^{A}(\mathbf{k},\omega)$ where each of them are the real-time Fourier transformation of the corresponding real-time thermal averaged Green's function. The thermal average is done in the grand canonical ensemble.

It is proved that we can express $\bar{G}(\mathbf{k},\omega)$ for Fermi system as \begin{equation} \bar{G}(\mathbf{k},\omega) = [1 + e^{-\beta \hbar \omega}]^{-1}{\bar{G}}^{R}(\mathbf{k},\omega)+[1+e^{\beta \hbar \omega}]^{-1} {\bar{G}}^{A}(\mathbf{k},\omega). \end{equation} This is in Eq. (31.24) of Fetter and I totally agree with this expression as I can really derive it. However, things get weird when I try to apply this expression to non-interacting Fermi system.

Suppose we have a non-interacting Fermi system with grand canonical Hamiltonian $K = \sum_{\mathbf{k}} \left( \epsilon_{\mathbf{k}} - \mu\right)c^{\dagger}_{\mathbf{k}}c_{\mathbf{k}} = \sum_{\mathbf{k}}\xi_{\mathbf{k}}c^{\dagger}_{\mathbf{k}}c_{\mathbf{k}}$ where $\epsilon_{\mathbf{k}}$ is the kinetic energy and $\mu$ is the chemical potential, then it is also known that \begin{equation} \bar{G}(\mathbf{k},\omega) = \frac{1}{1+e^{-\beta \xi_{\mathbf{k}}}}\cdot\frac{1}{\omega - \hbar^{-1}\xi_{\mathbf{k}}+i\eta} + \frac{1}{1+e^{\beta \xi_{\mathbf{k}}}}\cdot\frac{1}{\omega - \hbar^{-1}\xi_{\mathbf{k}} - i\eta}, \text{ (Eq. (31.38) in Fetter)} \end{equation} \begin{equation} \bar{G}^{R}(\mathbf{k},\omega) = \frac{1}{\omega - \hbar^{-1}\xi_{\mathbf{k}}+i\eta}, \end{equation} \begin{equation} \bar{G}^{A}(\mathbf{k},\omega) = \frac{1}{\omega - \hbar^{-1}\xi_{\mathbf{k}}-i\eta}, \end{equation} where we will take $\eta \to 0^{+}$ in the end. However, it is not obvious to me that \begin{equation} \bar{G}(\mathbf{k},\omega) = [1 + e^{-\beta \hbar \omega}]^{-1}{\bar{G}}^{R}(\mathbf{k},\omega)+[1+e^{\beta \hbar \omega}]^{-1} {\bar{G}}^{A}(\mathbf{k},\omega). \end{equation} can help us obtain the correct $\bar{G}(\mathbf{k},\omega)$. This is because if we just plug in the non-interacting $\bar{G}^{A}(\mathbf{k},\omega)$ and $\bar{G}^{R}(\mathbf{k},\omega)$ which I have shown above, we will obtain \begin{equation} \bar{G}(\mathbf{k},\omega) = \frac{1}{1 + e^{-\beta \hbar \omega}}\cdot\frac{1}{\omega - \hbar^{-1}\xi_{\mathbf{k}}+i\eta}+\frac{1}{1+e^{\beta \hbar \omega}}\cdot \frac{1}{\omega - \hbar^{-1}\xi_{\mathbf{k}}-i\eta} \end{equation} which is not (or at least that that obvious?) equal to \begin{equation} \bar{G}(\mathbf{k},\omega) = \frac{1}{1+e^{-\beta \xi_{\mathbf{k}}}}\cdot\frac{1}{\omega - \hbar^{-1}\xi_{\mathbf{k}}+i\eta} + \frac{1}{1+e^{\beta \xi_{\mathbf{k}}}}\cdot\frac{1}{\omega - \hbar^{-1}\xi_{\mathbf{k}} - i\eta} \text{ (Eq. (31.38) in Fetter)}. \end{equation} In the first expression we still have a $\omega$ dependence in $\frac{1}{1 + e^{-\beta \hbar \omega}}$ and $\frac{1}{1+e^{\beta \hbar \omega}}$ but in the second expression we have instead $\frac{1}{1+e^{-\beta \xi_{\mathbf{k}}}}$ and $\frac{1}{1+e^{\beta \xi_{\mathbf{k}}}}$. Are there suggestion to reconcile this? Thanks!


1 Answer 1


You are completely correct that the identity of the two equations is not trivial. Keep in mind however, that the Fourier-transform of a real-time-Green-function is indeed a distribution. Interpreting the r.h.s. in the distributional sense and using the Sokhotski-Plemelj-theorem helps to resolve your issues.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.