Computationally, there are various ways to calculate the electron-phonon coupling constants as a function of phonon index and momentum. My question is about the reverse process. Let us say that an experimentalist is (miraculously?) able to measure the effective Coulomb interaction in a material $V_{\textrm{eff}} (\mathbf{k}-\mathbf{k}', \nu)$ defined as:

$$V_{\textrm{eff}} (\mathbf{k}-\mathbf{k}', \nu) = \frac{V_0(\mathbf{k}-\mathbf{k}')}{\epsilon(\mathbf{k}-\mathbf{k}', \nu)}\tag{1}$$

where $V_0(\mathbf{k}-\mathbf{k}') = \frac{4\pi e^2}{\vert \mathbf{k}-\mathbf{k}' \vert^2}$ is the bare Coulomb interaction, and $\epsilon(\mathbf{k}-\mathbf{k}', \nu)$ is the dielectric function.

My question is the following: Is there a generic, model-independent way to extract out the electron-phonon coupling constants (or coupling functions, rather) or the $k$-resolved Eliashberg function from such a measurement of $V_{\textrm{eff}} (\mathbf{k}-\mathbf{k}', \nu)$ (or $\epsilon(\mathbf{k}-\mathbf{k}', \nu)$)?

Presumably, phonons should appear as poles in $\epsilon^{-1}(\mathbf{k}-\mathbf{k}', \nu)$, so clearly some information about phonons is contained in the dielectric function, but it is unclear to me what role electron-phonon coupling plays in the appearance of phonons in the dielectric function.

Edit: To be clear, I am not interested in comparing simulated dielectric functions (where coupling constants are inputs) to experiment dielectric functions. I am interested in whether the physics of electron-phonon coupling is fully contained and extractable from the dielectric function in principle.

  • $\begingroup$ +1. This is a nice question! KFC, I'd be grateful if you could take a look at my last edit for improved formatting of the LaTeX equations with Roman subscripts, and if you could replicate that throughout the rest of the question. This question is also a bit long, so might need a bounty to attract more interest! $\endgroup$ Dec 25, 2021 at 21:50
  • $\begingroup$ @Nike Dattani, thank you! I believe I had run a bounty on this question a few months ago but unfortunately did not get an answer. It is a quite odd question, so perhaps that's why. $\endgroup$
    – KFC
    Dec 26, 2021 at 5:42
  • 1
    $\begingroup$ Sorry to hear about that. I see that your bounty was added only 2 days after the question was posted (the minimum amount of wait time between posting and bounty). I was extremely busy during October and November, so probably missed it. In the future I'd recommend waiting a bit before posting a bounty, so that more people have a chance to see and think about the question (this site is new, and only has 3.7k users compared to 250k on Physics.SE and 16 million on StackOverflow), so you can't expect something to be seen right away. A second bounty may give this question sufficient attention. $\endgroup$ Dec 26, 2021 at 22:24

1 Answer 1


Your equation for the effective Coulomb interaction is given as Eq. 7.203 this PDF of Mahan's "Many-Particle Physics" book, but he also writes the same quantity in terms of the screened ("screening" it discussed in, e.g. Section 5.4 of the same book) electron-phonon interaction (in his notation):

$$ \tag{1} V_\textrm{eff} = \frac{V_q}{\varepsilon_\textrm{total}} = \frac{V_q^{(\infty)}}{\varepsilon(q,\omega)} + V_{\textrm{sc-ph}}(q,\omega), $$

in which,

$$\tag{2} V_q^{(\infty)} = \frac{V_q}{\varepsilon_\infty}.$$

Therefore if you obtain $V_\textrm{eff}$ and the $\varepsilon (q,\omega)$ empirically, then the screened electron-phonon coupling function (or perhaps, the Fourier transform of it) could be obtained by:

$$\tag{4} V_{\textrm{sc-ph}}(q,\omega) = V_\textrm{eff} - \frac{V_q}{\varepsilon_\infty \varepsilon(q,\omega)}. $$

If empirically we have only $\varepsilon (q,\omega)$, then we can get the difference between $V_{\textrm{sc-ph}}$ and $V_\textrm{eff}$, but not $V_{\textrm{sc-ph}}$ alone unless we're able to express $V_\textrm{eff}$ in terms of known/obtainable quantities in another way (a bit like the "two equations for two unknowns" concept).

If empirically we have only $V_\textrm{eff}$, then to get $V_{\textrm{sc-ph}}$ alone we would need to express $\varepsilon (q,\omega)$ in terms of known/obtainable quantities, which is discussed a bit in Section 5.5 of the same book by Mahan. An "exact" expression is given as:

        enter image description here

and a definition in terms of other objects (the impurity charge density $\rho_i(\mathbf{q})$ and screening charge density $\rho_s(\mathbf{q})$) is here:

enter image description here

and linear screening models simply assume:

enter image description here

The opening sentence of Section 5.5 of the same book by Mahan, says that even for the uniform electron gas, no one had derived an exact solution to Eq. 5.110 for the dielectric function. He presented four different approximate solutions: Thomas-Fermi (Section 5.5.1), Lindhard/RPA (Section 5.5.2), Hubbard (Section 5.5.3), Singwi-Sjoelander (Section 5.5.4) and that this is "only a small subset of the vast number which are available" (and that was decades ago).

The RPA is discussed again in Chapter 7, so I'll give the expression for $\varepsilon(q,\omega)$ from there:

enter image description here

For some cases the expressions can be spelled out more directly, for example if you have a polar solid insulator with conduction electrons, you would simply get:

enter image description here

which makes it easier to get your $V_{\textrm{sc-ph}}(q,\omega)$ function in terms from your empirical $V_\textrm{eff}$.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .