On p22 of "Green's Functions for Solid State Physicists" by Doniach and SondHeimer, there is the following definition:

$$G^0(\omega)=\frac{1}{2M\Omega_0}\left( \frac{1}{\omega-\Omega_0+i\eta} - \frac{1}{\omega+\Omega_0-i\eta} \right)$$

which (according to the author) simplifies to


for $\eta>0$. They then proceed to explain that the poles occur at $\omega=\pm (\Omega_0-i\eta)$. I don't understand this step. Moreover when I calculate the poles of the two different (but apparently equivalent) expressions with Mathematica I get

for the first one and

for the second one. What is going on here? Why are the poles different if these two expressions are equivalent. I guess maybe some approximation has been made because $\eta$ is an infinitesimal constant but ultimately I want to use this same technique with a finite $\eta$ corresponding to damping.

Any insight gratefully received :)

Edit 1: Applying these ideas to a finite (non-infinitesimal) $\eta$

Thanks to some comments and answers I understand the situation. However I still have some confusion which I'll explain here, but perhaps this should be a different question. In the book I reference, the authors explain that instead, if you have some finite value for $\eta$, suppose there is a pole at $\Omega_0-i\Gamma_k$ then the Fourier transform $G^0(t)$ acquires a real exponential factor $\exp(-\Gamma_k t)$ (whereas with an infinitesimal $\eta$ there is no factor left over in the Fourier transform, corresponding to an infinite lifetime). In my work I have derived the Green's function


For this particular $G_q(\omega)$ what will the poles be, and what will the exponential decay factor be? I'm assuming it won't simply be $\exp(-\omega\gamma t)$ because I can't play the same tricks as above because it's a finite, not infinitesimal factor?

  • 4
    $\begingroup$ the actual simplification is $G^0=\frac{1}{M\Omega}\frac{\Omega-i\eta}{\eta^2+\omega^2-\Omega^2+2i\eta\Omega}$. The $-i\eta$ factor in the numerator can be droped (its irrelevant for the poles). The $\eta^2$ can be dropped ("second order infinitesimal"). Finally (and most importantly), the factor $2i\eta\Omega$ is redefined into $i\tilde\eta$, where $\tilde\eta=\Omega\eta$, which has the same sign as $\eta$, and is also infinitesimal. In the end, the author drops the tilde (for notational convenience). $\endgroup$ Jan 1 '16 at 20:03

Your two expressions for $G^0$ are not identical, only when $\eta$ is taken to be a formal infinitesimal constant (so we take $\eta$ and $\text{const.} \eta$ to be the same and $\eta^2 \simeq 0$). This is useful if you only use $\eta$ to locate the zeroes of the denominator on the right place. In reality, the first formula can be rewritten as

$$ G^0(\omega)= \frac{1-i \eta/\Omega_0}{M(\omega^2 - \Omega_0^2 + 2 i \eta \Omega_0 + \eta^2)}.$$

Only if you ignore the $i \eta/\Omega_0$ term in the numerator and the $\eta^2$ term in the denominator to zero, and if you rescale $2 i \eta \Omega_0 \to i \eta$, you recover the second formula.


$\eta$ does not necessarily represent the same quantity in the two expressions. Since $\eta$ is a positive real infinitesimal, there is no meaningful difference between, say, $\eta$ and $\eta/|\Omega|$. If you keep this in mind and expand the square roots to linear order in $\eta$ (the only order that matters, since $\eta$ is infinitesimal), the results should agree.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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