SciPost Submission Page
Electric polarization and magnetization in metals
by P. T. Mahon, J. E. Sipe
This is not the latest submitted version.
This Submission thread is now published as
Submission summary
Authors (as registered SciPost users): | Perry Mahon |
Submission information | |
---|---|
Preprint Link: | scipost_202206_00031v1 (pdf) |
Date submitted: | 2022-06-27 19:28 |
Submitted by: | Mahon, Perry |
Submitted to: | SciPost Physics |
Ontological classification | |
---|---|
Academic field: | Physics |
Specialties: |
|
Approach: | Theoretical |
Abstract
A feature of the modern theory of polarization is that metallic systems do not admit a well-defined ground state polarization. This is predicated on the assumption that the electronic contribution to the polarization is well-defined if and only if the electronic ground state is "localized." If instead one takes the view that the electronic polarization is more fundamentally related to the existence of a complete set of exponentially localized Wannier functions, a definition is always admitted. This is the perspective we have adopted in the unified theory of microscopic polarization and magnetization fields that we have previously developed. Interestingly, when the modern theory admits a well-defined polarization, in particular for "trivial" insulators, these philosophically different approaches agree. Comparison with the modern theory of magnetization is somewhat different; we find agreement for the ground state orbital magnetization in "trivial" insulators and as well the predicted magnetoelectric effect, but disagree with later thermodynamic extensions to include metals and Chern insulators in that description. In addition, we also provide a novel perspective on the distinct contributions to the electrical conductivity tensor in the long-wavelength limit. In particular we find that, in the absence of any scattering mechanisms, the dc divergence of that tensor arises from what we identify as a free current density and the finite-frequency generalization of the anomalous Hall contribution arises from a combination of bound and free current densities.
Current status:
Reports on this Submission
Report #1 by Anonymous (Referee 3) on 2022-8-4 (Invited Report)
- Cite as: Anonymous, Report on arXiv:scipost_202206_00031v1, delivered 2022-08-04, doi: 10.21468/SciPost.Report.5494
Strengths
As detailed in my report some elements are missing or non sufficiently clear to be able to identify the strengths and weaknesses of this work.
Weaknesses
As detailed in my report some elements are missing or non sufficiently clear to be able to identify the strengths and weaknesses of this work.
Report
Dear editor,
The authors propose a new definition of the electric polarization and magnetization. The authors claim that, contrary to previous definitions of these quantities (in particular, in the modern theories of polarization and magnetization), the new definitions are also well-defined for metals. My main criticism of this work is the absence of any calculation (analytical or numerical) that would support this claim and also, more generally, that would confirm the viability of the authors’ approach for practical calculations. I note that also in other recent works by the same authors (PRB 99 235140, PRR 2 033126, PRR 2 043110) in which they present the same general method I have not found any calculation. Would it be possible to add some simple calculation, e.g. tight-binding or other, to illustrate the authors’ approach? As is well-known several complications might arise when putting a new method to the test.
Please find some further questions and comments below
1) The main results of the paper seem to be the definition of the polarization in Eq. (26) and the definition of the magnetization in Eq. (27). Is this indeed the case?
2) The only thing that can be measured in experiment is the change of polarization which is fully defined in terms of the time integral of the current density (assuming there is no magnetization). This works equally well for insulators and for metals. So, what is the importance of defining an absolute polarization which is a gauge-dependent quantity? (see also point 7 below)
3) In the abstract the authors write “If instead one takes the view that the electronic polarization is more fundamentally related to the existence of a complete set of exponentially localized Wannier functions, a definition is always admitted.” I do not see what is fundamental about the existence of a complete set of exponentially localized Wannier functions (ELWF). They might be very useful in practice for numerical calculations and/or for interpretation of the results but what is fundamental about them?
4) The authors use site-based quantities such as a site density and site current density but they are never explicitly defined. Could the authors please add these definitions (or point to the appropriate equations in earlier works)? Are the definitions general or are they contingent on the choice of using ELWF?
5) In the introduction it is written that a lattice sums are performed to give the microscopic polarization and magnetization. Lattice sums are generally conditionally convergent which could lead to convergence problems in practical calculations. Could the authors comment on this?
6) Also in the introduction, the authors write “One reason is that usual calculations made in minimal coupling can require the identification of sum rules to show properly behaved results at low frequencies - especially if nonlinear optical response is calculated, which is a future direction for this work - and that is not a difficulty with the calculations presented here, since the response is calculated as due to electric and magnetic fields.” In the linear response the problem of divergencies at low frequencies have been addressed in the literature and several solutions have been proposed (PRB 82 035104, PRB 95 155203). Indeed, I think the use of sum rules is an elegant approach to avoid divergencies.
7) Just below the authors write “A second reason is that in an insulator there is a clear physical significance to the response of the polarization to applied fields, as has been demonstrated within the “modern theory,"” Could the authors explain what this physical significance is?
8) Why do the authors neglect local-field corrections which can be very important? Moreover, no calculations are performed in this work, everything is purely formal, so why neglect them?
9) The authors write “Consideration of the phenomena related to spatially-varying electromagnetic fields is left for future work.” but one page further they write “The primary difference between our approach and those is that this investigation is a limiting case of a more general framework within which spatial and temporal variation of electric and magnetic fields can be into account; that is not the case in earlier works.” If this is the primary difference it would seem important to not defer spatially-varying electromagnetic fields for future work but to include them here.
10) In the conclusions the authors write “A positive feature of the approach implemented here is that it captures the complete finite-frequency response of a metallic crystal; there is no need to implement distinct approaches to calculate the anomalous Hall and Drude contributions to J(1).” Could the authors compare to PRB 86 125139 which also seems to include all contributions. Drude contributions are also included in PRB 74 245117 but not anomalous Hall contributions because time-reversal symmetry is assumed. However, I would expect that also this contribution is present if this assumption is removed.
11) In appendix B the authors write “Although most derivations [43] work with Bloch functions from the onset, some issues related to the position operator can be avoided if one works, at least initially, in the Hilbert space containing ELWFs, the space of square-integrable functions, where the usual position operator is well-defined.” The usual position operator is not compatible with periodic boundary conditions (PBC) and, therefore, ill-defined whenever PBC are applied. Therefore, I do not understand why the authors claim that by invoking ELWFs the usual position operator is well-defined. Could the authors explain this point in more detail? I note that recently a position operator has been defined that is compatible with PBC (PRB 99 205144, PRB 105 235201).
Some minor points
12) The authors write “However, often times, and as we will take to be the case in this paper, a number of Hilbert subbundles of the Bloch bundle are trivializable, …”. What is meant by “trivializable” ?
13) The conclusion section is almost two pages long. The authors might consider to be more concise.
14) References [25] and [60] are the same.
Claudio Attaccalite on 2022-09-05 [id 2789]
Dear authors
while I'm searching for an additional referee, I would like to ask you some questions on your manuscript, that you can answer at the same time as your answer to the referees
1) the use of Wannier Functions makes difficult the application to the electron dynamics, because it is difficult to generate ELWF for de-localized conduction bands. May the authors comment on this point. Moreover, is it possible to reformulate the present result with non-orthogonal localized basis set?
2) In Ref. 5 a generalization of Magnetization for metals and topological insulators is proposed, do the authors compared their results with that formula?
3) Is it not possible to recast all equations in the Bloch space? At the end there is an equivalence between Wannier and Bloch space. In this case it will much easy to see the difference with standard literature.
4) In the introduction the authors, say that it is difficult to take into account "variation of the optical fields over a unit cell". I do not see many difficulties to treat fields that are not uniform, in the Bloch formalism, it is sufficient that the momentum of the field to be compatible q=k-k', similar to what is done with phonons calculations. Even if I agree that a formalism in ELWFs is more general and could thread both cases q=0 and q/=0.
5) Does this formalism go again the Kohn argument of localization in an insulator? Is possible to extend the present ideas to a correlated system?
6) Is it possible to quantify in model or a simple calculation the differences with Modern Theory of Polarization in the case a Chern insulator ? is there a way to compare this formalism with experimental measurements?