qpms/lepaper/infinite.lyx

1274 lines
30 KiB
Plaintext
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

#LyX 2.4 created this file. For more info see https://www.lyx.org/
\lyxformat 584
\begin_document
\begin_header
\save_transient_properties true
\origin unavailable
\textclass article
\use_default_options true
\maintain_unincluded_children false
\language english
\language_package default
\inputencoding utf8
\fontencoding auto
\font_roman "default" "TeX Gyre Pagella"
\font_sans "default" "default"
\font_typewriter "default" "default"
\font_math "auto" "auto"
\font_default_family default
\use_non_tex_fonts false
\font_sc false
\font_roman_osf true
\font_sans_osf false
\font_typewriter_osf false
\font_sf_scale 100 100
\font_tt_scale 100 100
\use_microtype false
\use_dash_ligatures true
\graphics default
\default_output_format default
\output_sync 0
\bibtex_command default
\index_command default
\float_placement class
\float_alignment class
\paperfontsize default
\spacing single
\use_hyperref true
\pdf_author "Marek Nečada"
\pdf_bookmarks true
\pdf_bookmarksnumbered false
\pdf_bookmarksopen false
\pdf_bookmarksopenlevel 1
\pdf_breaklinks false
\pdf_pdfborder false
\pdf_colorlinks false
\pdf_backref false
\pdf_pdfusetitle true
\papersize default
\use_geometry false
\use_package amsmath 1
\use_package amssymb 1
\use_package cancel 1
\use_package esint 1
\use_package mathdots 1
\use_package mathtools 1
\use_package mhchem 1
\use_package stackrel 1
\use_package stmaryrd 1
\use_package undertilde 1
\cite_engine basic
\cite_engine_type default
\biblio_style plain
\use_bibtopic false
\use_indices false
\paperorientation portrait
\suppress_date false
\justification true
\use_refstyle 1
\use_minted 0
\use_lineno 0
\index Index
\shortcut idx
\color #008000
\end_index
\secnumdepth 3
\tocdepth 3
\paragraph_separation indent
\paragraph_indentation default
\is_math_indent 0
\math_numbering_side default
\quotes_style english
\dynamic_quotes 0
\papercolumns 1
\papersides 1
\paperpagestyle default
\tablestyle default
\tracking_changes false
\output_changes false
\html_math_output 0
\html_css_as_file 0
\html_be_strict false
\end_header
\begin_body
\begin_layout Section
Infinite periodic systems
\begin_inset CommandInset label
LatexCommand label
name "sec:Infinite"
\end_inset
\begin_inset FormulaMacro
\newcommand{\dlv}{\vect a}
\end_inset
\begin_inset FormulaMacro
\newcommand{\rlv}{\vect b}
\end_inset
\end_layout
\begin_layout Standard
Although large finite systems are where MSTMM excels the most, there are
several reasons that makes its extension to infinite lattices (where periodic
boundary conditions might be applied) desirable as well.
Other methods might be already fast enough, but MSTMM will be faster in
most cases in which there is enough spacing between the neighboring particles.
MSTMM works well with any space group symmetry the system might have (as
opposed to, for example, FDTD with a cubic mesh applied to a honeycomb
lattice), which makes e.g.
application of group theory in mode analysis quite easy.
\begin_inset Note Note
status open
\begin_layout Plain Layout
Topology anoyne?
\end_layout
\end_inset
And finally, having a method that handles well both infinite and large
finite systems gives a possibility to study finite-size effects in periodic
scatterer arrays.
\end_layout
\begin_layout Subsection
Formulation of the problem
\end_layout
\begin_layout Standard
Let us have a linear system of compact EM scatterers on a homogeneous background
as in Section
\begin_inset CommandInset ref
LatexCommand eqref
reference "subsec:Multiple-scattering"
plural "false"
caps "false"
noprefix "false"
\end_inset
, but this time, the system shall be periodic: let there be a
\begin_inset Formula $d$
\end_inset
-dimensional (
\begin_inset Formula $d$
\end_inset
can be 1, 2 or 3) Bravais lattice embedded into the three-dimensional real
space, with lattice vectors
\begin_inset Formula $\left\{ \dlv_{i}\right\} _{i=1}^{d}$
\end_inset
, and let the lattice points be labeled with an
\begin_inset Formula $d$
\end_inset
-dimensional integer multi-index
\begin_inset Formula $\vect n\in\ints^{d}$
\end_inset
, so the lattice points have cartesian coordinates
\begin_inset Formula $\vect R_{\vect n}=\sum_{i=1}^{d}n_{i}\vect a_{i}$
\end_inset
.
There can be several scatterers per unit cell with indices
\begin_inset Formula $\alpha$
\end_inset
from set
\begin_inset Formula $\mathcal{P}_{1}$
\end_inset
and (relative) positions inside the unit cell
\begin_inset Formula $\vect r_{\alpha}$
\end_inset
; any particle of the periodic system can thus be labeled by a multiindex
from
\begin_inset Formula $\mathcal{P}=\ints^{d}\times\mathcal{P}_{1}$
\end_inset
.
The scatterers are located at positions
\begin_inset Formula $\vect r_{\vect n,\alpha}=\vect R_{\vect n}+\vect r_{\alpha}$
\end_inset
and their
\begin_inset Formula $T$
\end_inset
-matrices are periodic,
\begin_inset Formula $T_{\vect n,\alpha}=T_{\alpha}$
\end_inset
.
In such system, the multiple-scattering problem
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Multiple-scattering problem"
plural "false"
caps "false"
noprefix "false"
\end_inset
can be rewritten as
\end_layout
\begin_layout Standard
\begin_inset Formula
\begin{equation}
\outcoeffp{\vect n,\alpha}-T_{\alpha}\sum_{\left(\vect m,\beta\right)\in\mathcal{P}\backslash\left\{ \left(\vect n,\alpha\right)\right\} }\tropsp{\vect n,\alpha}{\vect m,\beta}\outcoeffp{\vect m,\beta}=T_{\alpha}\rcoeffincp{\vect n,\alpha}.\quad\left(\vect n,\alpha\right)\in\mathcal{P}\label{eq:Multiple-scattering problem periodic}
\end{equation}
\end_inset
\end_layout
\begin_layout Standard
Due to periodicity, we can also write
\begin_inset Formula $\tropsp{\vect n,\alpha}{\vect m,\beta}=\tropsp{\alpha}{\beta}\left(\vect R_{\vect m}-\vect R_{\vect n}\right)=\tropsp{\alpha}{\beta}\left(\vect R_{\vect m-\vect n}\right)=\tropsp{\vect 0,\alpha}{\vect m-\vect n,\beta}$
\end_inset
.
Assuming quasi-periodic right-hand side with quasi-momentum
\begin_inset Formula $\vect k$
\end_inset
,
\begin_inset Formula $\rcoeffincp{\vect n,\alpha}=\rcoeffincp{\vect 0,\alpha}\left(\vect k\right)e^{i\vect k\cdot\vect R_{\vect n}}$
\end_inset
, the solutions of
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Multiple-scattering problem periodic"
plural "false"
caps "false"
noprefix "false"
\end_inset
will be also quasi-periodic according to Bloch theorem,
\begin_inset Formula $\outcoeffp{\vect n,\alpha}=\outcoeffp{\vect 0,\alpha}\left(\vect k\right)e^{i\vect k\cdot\vect R_{\vect n}}$
\end_inset
, and eq.
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Multiple-scattering problem periodic"
plural "false"
caps "false"
noprefix "false"
\end_inset
can be rewritten as follows
\begin_inset Formula
\begin{align}
\outcoeffp{\vect 0,\alpha}\left(\vect k\right)e^{i\vect k\cdot\vect R_{\vect n}}-T_{\alpha}\sum_{\left(\vect m,\beta\right)\in\mathcal{P}\backslash\left\{ \left(\vect n,\alpha\right)\right\} }\tropsp{\vect n,\alpha}{\vect m,\beta}\outcoeffp{\vect 0,\beta}\left(\vect k\right)e^{i\vect k\cdot\vect R_{\vect m}} & =T_{\alpha}\rcoeffincp{\vect 0,\alpha}\left(\vect k\right)e^{i\vect k\cdot\vect R_{\vect n}},\nonumber \\
\outcoeffp{\vect 0,\alpha}\left(\vect k\right)-T_{\alpha}\sum_{\left(\vect m,\beta\right)\in\mathcal{P}\backslash\left\{ \left(\vect n,\alpha\right)\right\} }\tropsp{\vect 0,\alpha}{\vect m-\vect n,\beta}\outcoeffp{\vect 0,\beta}\left(\vect k\right)e^{i\vect k\cdot\vect R_{\vect m-\vect n}} & =T_{\alpha}\rcoeffincp{\vect 0,\alpha}\left(\vect k\right),\nonumber \\
\outcoeffp{\vect 0,\alpha}\left(\vect k\right)-T_{\alpha}\sum_{\left(\vect m,\beta\right)\in\mathcal{P}\backslash\left\{ \left(\vect 0,\alpha\right)\right\} }\tropsp{\vect 0,\alpha}{\vect m,\beta}\outcoeffp{\vect 0,\beta}\left(\vect k\right)e^{i\vect k\cdot\vect R_{\vect m}} & =T_{\alpha}\rcoeffincp{\vect 0,\alpha}\left(\vect k\right),\nonumber \\
\outcoeffp{\vect 0,\alpha}\left(\vect k\right)-T_{\alpha}\sum_{\beta\in\mathcal{P}}W_{\alpha\beta}\left(\vect k\right)\outcoeffp{\vect 0,\beta}\left(\vect k\right) & =T_{\alpha}\rcoeffincp{\vect 0,\alpha}\left(\vect k\right),\label{eq:Multiple-scattering problem unit cell}
\end{align}
\end_inset
so we reduced the initial scattering problem to one involving only the field
expansion coefficients from a single unit cell, but we need to compute
the
\begin_inset Quotes eld
\end_inset
lattice Fourier transform
\begin_inset Quotes erd
\end_inset
of the translation operator,
\begin_inset Formula
\begin{equation}
W_{\alpha\beta}(\vect k)\equiv\sum_{\vect m\in\ints^{d}}\left(1-\delta_{\alpha\beta}\delta_{\vect m\vect 0}\right)\tropsp{\vect 0,\alpha}{\vect m,\beta}e^{i\vect k\cdot\vect R_{\vect m}},\label{eq:W definition}
\end{equation}
\end_inset
evaluation of which is possible but quite non-trivial due to the infinite
lattice sum, so we explain it separately in Sect.
\begin_inset CommandInset ref
LatexCommand eqref
reference "subsec:W operator evaluation"
plural "false"
caps "false"
noprefix "false"
\end_inset
.
\end_layout
\begin_layout Standard
As in the case of a finite system, eq.
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Multiple-scattering problem unit cell"
plural "false"
caps "false"
noprefix "false"
\end_inset
can be written in a shorter block-matrix form,
\begin_inset Formula
\begin{equation}
\left(I-TW\right)\outcoeffp{\vect 0}\left(\vect k\right)=T\rcoeffincp{\vect 0}\left(\vect k\right)\label{eq:Multiple-scattering problem unit cell block form}
\end{equation}
\end_inset
Eq.
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Multiple-scattering problem unit cell"
plural "false"
caps "false"
noprefix "false"
\end_inset
can be used to calculate electromagnetic response of the structure to external
quasiperiodic driving field most notably a plane wave.
However, the non-trivial solutions of the equation with right hand side
(i.e.
the external driving) set to zero,
\begin_inset Formula
\begin{equation}
\left(I-TW\right)\outcoeffp{\vect 0}\left(\vect k\right)=0,\label{eq:lattice mode equation}
\end{equation}
\end_inset
describes the
\emph on
lattice modes.
\emph default
Non-trivial solutions to
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:lattice mode equation"
plural "false"
caps "false"
noprefix "false"
\end_inset
exist if the matrix on the left-hand side
\begin_inset Formula $M\left(\omega,\vect k\right)=\left(I-T\left(\omega\right)W\left(\omega,\vect k\right)\right)$
\end_inset
is singular this condition gives the
\emph on
dispersion relation
\emph default
for the periodic structure.
Note that in realistic (lossy) systems, at least one of the pair
\begin_inset Formula $\omega,\vect k$
\end_inset
will acquire complex values.
The solution
\begin_inset Formula $\outcoeffp{\vect 0}\left(\vect k\right)$
\end_inset
is then obtained as the right
\begin_inset Note Note
status open
\begin_layout Plain Layout
CHECK!
\end_layout
\end_inset
singular vector of
\begin_inset Formula $M\left(\omega,\vect k\right)$
\end_inset
corresponding to the zero singular value.
\end_layout
\begin_layout Subsection
Numerical solution
\end_layout
\begin_layout Standard
In practice, equation
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Multiple-scattering problem unit cell block form"
plural "false"
caps "false"
noprefix "false"
\end_inset
is solved in the same way as eq.
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Multiple-scattering problem block form"
plural "false"
caps "false"
noprefix "false"
\end_inset
in the multipole degree truncated form.
\end_layout
\begin_layout Standard
The lattice mode problem
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:lattice mode equation"
plural "false"
caps "false"
noprefix "false"
\end_inset
is (after multipole degree truncation) solved by finding
\begin_inset Formula $\omega,\vect k$
\end_inset
for which the matrix
\begin_inset Formula $M\left(\omega,\vect k\right)$
\end_inset
has a zero singular value.
A naïve approach to do that is to sample a volume with a grid in the
\begin_inset Formula $\left(\omega,\vect k\right)$
\end_inset
space, performing a singular value decomposition of
\begin_inset Formula $M\left(\omega,\vect k\right)$
\end_inset
at each point and finding where the lowest singular value of
\begin_inset Formula $M\left(\omega,\vect k\right)$
\end_inset
is close enough to zero.
However, this approach is quite expensive, since
\begin_inset Formula $W\left(\omega,\vect k\right)$
\end_inset
has to be evaluated for each
\begin_inset Formula $\omega,\vect k$
\end_inset
pair separately (unlike the original finite case
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Multiple-scattering problem block form"
plural "false"
caps "false"
noprefix "false"
\end_inset
translation operator
\begin_inset Formula $\trops$
\end_inset
, which, for a given geometry, depends only on frequency).
Therefore, a much more efficient but not completely robust approach to
determine the photonic bands is to sample the
\begin_inset Formula $\vect k$
\end_inset
-space (a whole Brillouin zone or its part) and for each fixed
\begin_inset Formula $\vect k$
\end_inset
to find a corresponding frequency
\begin_inset Formula $\omega$
\end_inset
with zero singular value of
\begin_inset Formula $M\left(\omega,\vect k\right)$
\end_inset
using a minimisation algorithm (two- or one-dimensional, depending on whether
one needs the exact complex-valued
\begin_inset Formula $\omega$
\end_inset
or whether the its real-valued approximation is satisfactory).
Typically, a good initial guess for
\begin_inset Formula $\omega\left(\vect k\right)$
\end_inset
is obtained from the empty lattice approximation,
\begin_inset Formula $\left|\vect k\right|=\sqrt{\epsilon\mu}\omega/c_{0}$
\end_inset
(modulo reciprocal lattice points
\begin_inset Note Note
status open
\begin_layout Plain Layout
TODO write this in a clean way
\end_layout
\end_inset
).
A somehow challenging step is to distinguish the different bands that can
all be very close to the empty lattice approximation, especially if the
particles in the system are small.
In high-symmetry points of the Brilloin zone, this can be solved by factorising
\begin_inset Formula $M\left(\omega,\vect k\right)$
\end_inset
into irreducible representations
\begin_inset Formula $\Gamma_{i}$
\end_inset
and performing the minimisation in each irrep separately, cf.
Section
\begin_inset CommandInset ref
LatexCommand ref
reference "sec:Symmetries"
plural "false"
caps "false"
noprefix "false"
\end_inset
, and using the different
\begin_inset Formula $\omega_{\Gamma_{i}}\left(\vect k\right)$
\end_inset
to obtain the initial guesses for the nearby points
\begin_inset Formula $\vect k+\delta\vect k$
\end_inset
.
\end_layout
\begin_layout Standard
An alternative, more robust approach to generic minimisation algorithms
is Beyn's contour integral method
\begin_inset CommandInset citation
LatexCommand cite
key "beyn_integral_2012"
literal "false"
\end_inset
which finds the roots of
\begin_inset Formula $M\left(\omega,\vect k\right)=0$
\end_inset
inside an area enclosed by a given complex frequency plane contour, assuming
that
\begin_inset Formula $M\left(\omega,\vect k\right)$
\end_inset
is an analytical function of
\begin_inset Formula $\omega$
\end_inset
inside the contour.
A necessary prerequisite for this is that all the ingredients of
\begin_inset Formula $M\left(\omega,\vect k\right)$
\end_inset
are analytical as well.
It practice, this usually means that interpolation cannot be used in a
straightforward way for material properties or
\begin_inset Formula $T$
\end_inset
-matrices.
For material response, constant permittivity or Drude-Lorentz models suit
this purpose well.
The need to evaluate the
\begin_inset Formula $T$
\end_inset
-matrices precisely (without the speedup provided by interpolation) at many
points might cause a performance bottleneck for scatterers with more complicate
d shapes.
And finally, the integration contour has to evade any branch cuts appearing
in the lattice-summed translation operator
\begin_inset Formula $W\left(\omega,\vect k\right)$
\end_inset
, as described in the following.
\end_layout
\begin_layout Subsection
Computing the lattice sum of the translation operator
\begin_inset CommandInset label
LatexCommand label
name "subsec:W operator evaluation"
\end_inset
\end_layout
\begin_layout Standard
The problem in evaluating
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:W definition"
\end_inset
is the asymptotic behaviour of the translation operator,
\begin_inset Formula $\tropsp{\vect 0,\alpha}{\vect m,\beta}\sim\left|\vect R_{\vect m}\right|^{-1}e^{i\kappa\left|\vect R_{\vect m}\right|}$
\end_inset
, so that its lattice sum does not in the strict sense converge for any
\begin_inset Formula $d>1$
\end_inset
-dimensional lattice unless
\begin_inset Formula $\Im\kappa>0$
\end_inset
.
\begin_inset Note Note
status open
\begin_layout Plain Layout
\begin_inset Foot
status open
\begin_layout Plain Layout
Note that
\begin_inset Formula $d$
\end_inset
here is dimensionality of the lattice, not the space it lies in, which
I for certain reasons assume to be three.
(TODO few notes on integration and reciprocal lattices in some appendix)
\end_layout
\end_inset
\end_layout
\end_inset
The problem of poorly converging lattice sums has been originally solved
for electrostatic potentials with Ewald summation
\begin_inset CommandInset citation
LatexCommand cite
key "ewald_berechnung_1921"
literal "false"
\end_inset
.
Its basic idea is to decomposethe lattice-summed function in two parts:
a short-range part that decays fast and can be summed directly, and a long-rang
e part which decays poorly but is fairly smooth everywhere, so that its
Fourier transform decays fast enough, and to deal with the long range part
by Poisson summation over the reciprocal lattice.
The same idea can be used also in the case of linear electrodynamic problems,
just the technical details are more complicated than in electrostatics.
\end_layout
\begin_layout Standard
In eq.
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:translation operator singular"
plural "false"
caps "false"
noprefix "false"
\end_inset
we demonstratively expressed the translation operator elements as linear
combinations of (outgoing) scalar spherical wavefunctions
\begin_inset Formula $\sswfoutlm lm\left(\vect r\right)=h_{l}^{\left(1\right)}\left(r\right)\ush lm\left(\uvec r\right)$
\end_inset
, because for them, fortunately, exponentially convergent Ewald-type summation
formulae have been already developed
\begin_inset Note Note
status open
\begin_layout Plain Layout
add refs
\end_layout
\end_inset
\begin_inset CommandInset citation
LatexCommand cite
key "moroz_quasi-periodic_2006,linton_one-_2009,linton_lattice_2010"
literal "false"
\end_inset
and can be applied to our case.
If we formally label
\begin_inset Note Note
status open
\begin_layout Plain Layout
\begin_inset Marginal
status open
\begin_layout Plain Layout
FP: Check signs.
\end_layout
\end_inset
\end_layout
\end_inset
\begin_inset Formula
\begin{equation}
\sigma_{l,m}\left(\vect k,\vect s\right)=\sum_{\vect n\in\ints^{d}}\left(1-\delta_{\vect{R_{n}},\vect s}\right)e^{i\vect{\vect k}\cdot\left(\vect R_{\vect n}-\vect s\right)}\sswfoutlm lm\left(\vect{R_{n}}-\vect s\right),\label{eq:sigma lattice sums}
\end{equation}
\end_inset
we see from eqs.
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:translation operator singular"
plural "false"
caps "false"
noprefix "false"
\end_inset
,
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:W definition"
plural "false"
caps "false"
noprefix "false"
\end_inset
that the matrix elements of
\begin_inset Formula $W_{\alpha\beta}(\vect k)$
\end_inset
read
\begin_inset Note Note
status open
\begin_layout Plain Layout
\begin_inset Formula
\[
W_{\alpha\beta}(\vect k)\equiv\sum_{\vect m\in\ints^{d}}\left(1-\delta_{\alpha\beta}\right)\tropsp{\vect 0,\alpha}{\vect m,\beta}e^{i\vect k\cdot\vect R_{\vect m}},
\]
\end_inset
\end_layout
\end_inset
\begin_inset Formula
\begin{align*}
W_{\alpha,\tau lm;\beta,\tau l'm'}(\vect k) & =\sum_{\lambda=\left|l-l'\right|}^{l+l'}C_{lm;l'm'}^{\lambda}\sigma_{\lambda,m-m'}\left(\vect k,\vect r_{\beta}-\vect r_{\alpha}\right),\\
W_{\alpha,\tau lm;\beta,\tau'l'm'}(\vect k) & =\sum_{\lambda=\left|l-l'\right|+1}^{l+l'}D_{lm;l'm'}^{\lambda}\sigma_{\lambda,m-m'}\left(\vect k,\vect r_{\beta}-\vect r_{\alpha}\right),\quad\tau'\ne\tau,
\end{align*}
\end_inset
\begin_inset Marginal
status open
\begin_layout Plain Layout
Check signs
\end_layout
\end_inset
where the constant factors are exactly the same as in
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:translation operator constant factors"
plural "false"
caps "false"
noprefix "false"
\end_inset
.
\end_layout
\begin_layout Standard
For reader's reference, we list the Ewald-type formulae for lattice sums
\begin_inset Formula $\sigma_{l,m}\left(\vect k,\vect s\right)$
\end_inset
from
\begin_inset CommandInset citation
LatexCommand cite
key "linton_lattice_2010"
literal "false"
\end_inset
rewritten in a way that is independent on particular phase or normalisation
conventions of vector spherical harmonics.
\end_layout
\begin_layout Standard
In all three dimensionality cases, the lattice sums are divided into short-range
and long-range parts,
\begin_inset Formula $\sigma_{l,m}\left(\vect k,\vect s\right)=\sigma_{l,m}^{\left(\mathrm{S},\eta\right)}\left(\vect k,\vect s\right)+\sigma_{l,m}^{\left(\mathrm{L},\eta\right)}\left(\vect k,\vect s\right)$
\end_inset
depending on a positive parameter
\begin_inset Formula $\eta$
\end_inset
.
The short-range part has in all three cases the same form:
\begin_inset Note Note
status open
\begin_layout Plain Layout
FP: Check sign of s
\end_layout
\end_inset
\end_layout
\begin_layout Standard
\begin_inset Formula
\begin{multline}
\sigma_{l,m}^{\left(\mathrm{S},\eta\right)}\left(\vect k,\vect s\right)=-\frac{2^{l+1}i}{\kappa^{l+1}\sqrt{\pi}}\sum_{\vect n\in\ints^{d}}\left(1-\delta_{\vect{R_{n}},-\vect s}\right)\left|\vect{R_{n}+\vect s}\right|^{l}\ush lm\left(\vect{R_{n}+\vect s}\right)e^{i\vect k\cdot\left(\vect{R_{n}+\vect s}\right)}\\
\times\int_{\eta}^{\infty}e^{-\left|\vect{R_{n}+\vect s}\right|^{2}\xi^{2}}e^{-\kappa/4\xi^{2}}\xi^{2l}\ud\xi\\
+\delta_{\vect{R_{n}},-\vect s}\frac{\delta_{l0}\delta_{m0}}{\sqrt{4\pi}}\Gamma\left(-\frac{1}{2},-\frac{\kappa}{4\eta^{2}}\right)\ush lm\left(\vect{R_{n}+\vect s}\right).\label{eq:Ewald in 3D short-range part}
\end{multline}
\end_inset
Here
\begin_inset Formula $\Gamma(a,z)$
\end_inset
is the incomplete Gamma function.
The last (
\begin_inset Quotes eld
\end_inset
self-interaction
\begin_inset Quotes erd
\end_inset
) term in
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Ewald in 3D short-range part"
plural "false"
caps "false"
noprefix "false"
\end_inset
, which appears only when the displacement vector
\begin_inset Formula $\vect s$
\end_inset
coincides with a lattice point, is often noted separately in the literature.
\begin_inset Note Note
status open
\begin_layout Plain Layout
Poznámka ohledně zahození radiální části u kulových fcí?
\end_layout
\end_inset
The long-range part for cases
\begin_inset Formula $d=1,2$
\end_inset
reads
\begin_inset Note Note
status open
\begin_layout Plain Layout
FP: check sign of
\begin_inset Formula $\vect k$
\end_inset
\end_layout
\end_inset
\begin_inset Formula
\begin{multline}
\sigma_{l,m}^{\left(\mathrm{L},\eta\right)}\left(\vect k,\vect s\right)=-\frac{i^{l+1}}{\kappa^{d}\mathcal{A}}\pi^{2+\left(3-d\right)/2}2\left(\left(l-m\right)/2\right)!\left(\left(l+m\right)/2\right)!\times\\
\times\sum_{\vect K\in\Lambda^{*}}\ush lm\left(\vect k+\vect K\right)\sum_{j=0}^{\left[\left(l-\left|m\right|/2\right)\right]}\frac{\left(-1\right)^{j}\left(\left|\vect k+\vect K\right|/\kappa\right)^{l-2j}\Gamma\left(\frac{d-1}{2}-j,\frac{\kappa^{2}\gamma\left(\left|\vect k+\vect K\right|/\kappa\right)}{4\eta^{2}}\right)}{j!\left(\frac{1}{2}\left(l-m\right)-j\right)!\left(\frac{1}{2}\left(l+m\right)-j\right)!}\left(\gamma\left(\left|\vect k+\vect K\right|/\kappa\right)\right)^{2j+3-d}\label{eq:Ewald in 3D long-range part 1D 2D}
\end{multline}
\end_inset
and for
\begin_inset Formula $d=3$
\end_inset
\begin_inset Formula
\begin{equation}
\sigma_{l,m}^{\left(\mathrm{L},\eta\right)}\left(\vect k,\vect s\right)=\frac{4\pi i^{l+1}}{\kappa\mathcal{A}}\sum_{\vect K\in\Lambda^{*}}\frac{\left(\left|\vect k+\vect K\right|/\kappa\right)^{l}}{\kappa^{2}-\left|\vect k+\vect K\right|^{2}}e^{\left(\kappa^{2}-\left|\vect k+\vect K\right|^{2}\right)/4\eta^{2}}\ush lm\left(\vect k+\vect K\right).\label{eq:Ewald in 3D long-range part 3D}
\end{equation}
\end_inset
Here
\begin_inset Formula $\mathcal{A}$
\end_inset
is the unit cell volume (or length/area in the 1D/2D lattice cases).
The sums are taken over the reciprocal lattice
\begin_inset Formula $\Lambda^{*}$
\end_inset
with lattice vectors
\begin_inset Formula $\left\{ \vect b_{i}\right\} _{i=1}^{d}$
\end_inset
satisfying
\begin_inset Formula $\vect a_{i}\cdot\vect b_{j}=\delta_{ij}$
\end_inset
.
The function
\begin_inset Formula $\gamma\left(z\right)$
\end_inset
used in
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Ewald in 3D long-range part 1D 2D"
plural "false"
caps "false"
noprefix "false"
\end_inset
is defined as
\begin_inset Formula
\[
\gamma\left(z\right)=\left(z-1\right)^{\frac{1}{2}}\left(z+1\right)^{\frac{1}{2}},\quad-\frac{3\pi}{2}<\arg\left(z-1\right)<\frac{\pi}{2},-\frac{\pi}{2}<\arg\left(z+1\right)<\frac{3\pi}{2}.
\]
\end_inset
The Ewald parameter
\begin_inset Formula $\eta$
\end_inset
determines the pace of convergence of both parts.
The larger
\begin_inset Formula $\eta$
\end_inset
is, the faster
\begin_inset Formula $\sigma_{l,m}^{\left(\mathrm{S},\eta\right)}\left(\vect k,\vect s\right)$
\end_inset
converges but the slower
\begin_inset Formula $\sigma_{l,m}^{\left(L,\eta\right)}\left(\vect k,\vect s\right)$
\end_inset
converges.
Therefore (based on the lattice geometry) it has to be adjusted in a way
that a reasonable amount of terms needs to be evaluated numerically from
both
\begin_inset Formula $\sigma_{l,m}^{\left(\mathrm{S},\eta\right)}\left(\vect k,\vect s\right)$
\end_inset
and
\begin_inset Formula $\sigma_{l,m}^{\left(\mathrm{L},\eta\right)}\left(\vect k,\vect s\right)$
\end_inset
.
For one-dimensional, square, and cubic lattices, the optimal choice is
\begin_inset Formula $\eta=\sqrt{\pi}/p$
\end_inset
where
\begin_inset Formula $p$
\end_inset
is the direct lattice period
\begin_inset CommandInset citation
LatexCommand cite
key "linton_lattice_2010"
literal "false"
\end_inset
.
\begin_inset Note Note
status open
\begin_layout Plain Layout
\begin_inset Marginal
status open
\begin_layout Plain Layout
Whatabout different geometries?
\end_layout
\end_inset
\end_layout
\end_inset
\begin_inset Note Note
status open
\begin_layout Plain Layout
\begin_inset Marginal
status open
\begin_layout Plain Layout
FP: I have some error estimates derived in my notes.
Should I include them?
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout Standard
For a two-dimensional lattice, the incomplete
\begin_inset Formula $\Gamma$
\end_inset
-function
\begin_inset Formula $\Gamma\left(\frac{1}{2}-j,z\right)$
\end_inset
in the long-range part has a branch point at
\begin_inset Formula $z=0$
\end_inset
and special care has to be taken when choosing the appropriate branch.
If the wavenumber of the medium has a positive imaginary part,
\begin_inset Formula $\Im\kappa>0$
\end_inset
, then the translation operator elements
\begin_inset Formula $\trops_{\tau lm;\tau'l'm}\left(\kappa\vect r\right)$
\end_inset
decay exponentially as
\begin_inset Formula $\left|\vect r\right|\to\infty$
\end_inset
and the lattice sum in
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:W definition"
plural "false"
caps "false"
noprefix "false"
\end_inset
converges absolutely even in the direct space, and it is equal to the Ewald
sum with the principal value of the incomplete
\begin_inset Formula $\Gamma$
\end_inset
function being used in
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Ewald in 3D long-range part 1D 2D"
plural "false"
caps "false"
noprefix "false"
\end_inset
.
For other values of
\begin_inset Formula $\kappa$
\end_inset
, the branch choice is made in such way that
\begin_inset Formula $W_{\alpha\beta}\left(\vect k\right)$
\end_inset
is analytically continued even when the wavenumber's imaginary part crosses
the real axis.
The principal value of
\begin_inset Formula $\Gamma\left(\frac{1}{2}-j,z\right)$
\end_inset
has a branch cut at the negative real half-axis, which, considering the
lattice sum as a function of
\begin_inset Formula $\kappa$
\end_inset
, translates into branch cuts starting at
\begin_inset Formula $\kappa=\left|\vect k+\vect K\right|$
\end_inset
and continuing in straight lines towards
\begin_inset Formula $+\infty$
\end_inset
.
Therefore, in the quadrant
\begin_inset Formula $\Re z<0,\Im z\ge0$
\end_inset
we use the continuation of the principal value from
\begin_inset Formula $\Re z<0,\Im z<0$
\end_inset
instead of the principal branch
\begin_inset CommandInset citation
LatexCommand cite
after "8.2.9"
key "NIST:DLMF"
literal "false"
\end_inset
, moving the branch cut in the
\begin_inset Formula $z$
\end_inset
variable to the positive imaginary half-axis.
This moves the branch cuts w.r.t.
\begin_inset Formula $\kappa$
\end_inset
away from the real axis.
\begin_inset Note Note
status open
\begin_layout Plain Layout
TODO Figure.
\end_layout
\end_inset
Detailed physical interpretation of the remaining branch cuts is an open
question.
\begin_inset Note Note
status open
\begin_layout Plain Layout
Generally, a good choice for
\begin_inset Formula $\eta$
\end_inset
is TODO; in order to achieve accuracy TODO, one has to evaluate the terms
on TODO lattice points.
\end_layout
\end_inset
\begin_inset Note Note
status open
\begin_layout Plain Layout
(I HAVE SOME DERIVATIONS OF THE ESTIMATES IN MY NOTES; SHOULD I INCLUDE
THEM?)
\end_layout
\end_inset
\end_layout
\begin_layout Standard
In practice, the integrals in
\begin_inset CommandInset ref
LatexCommand eqref
reference "eq:Ewald in 3D short-range part"
plural "false"
caps "false"
noprefix "false"
\end_inset
can be easily evaluated by numerical quadrature and the incomplete
\begin_inset Formula $\Gamma$
\end_inset
-functions using the series 8.7.3 from
\begin_inset CommandInset citation
LatexCommand cite
key "NIST:DLMF"
literal "false"
\end_inset
.
\end_layout
\end_body
\end_document