Content deleted Content added
Citation bot (talk | contribs) Altered template type. Add: doi-access, article-number, pmid. Removed parameters. Some additions/deletions were parameter name changes. | Use this bot. Report bugs. | Suggested by AManWithNoPlan | Category:CS1 maint: unflagged free DOI | #UCB_Category 4/42 |
Swasthyapro (talk | contribs) Link suggestions feature: 3 links added. |
||
| (21 intermediate revisions by 3 users not shown) | |||
Line 58:
=== Dyadic Green’s function ===
:<math>
\mathbf{G}(\mathbf{r}) = \left[ \nabla \nabla + k^2 \mathbf{I} \right] \frac{e^{ikr}}{r},
</math>
where <math>k</math> is the [[wavenumber]], <math>\mathbf{I}</math> is the [[identity matrix]], and <math>\mathbf{r}</math> is the vector from the source dipole to the observation point. Evaluating the derivatives leads to the explicit form:
:<math>
\mathbf{G}(\mathbf{r}) = \frac{e^{ikr}}{r^3} \left[
k^2 r^2 \left( \mathbf{I} - \hat{\mathbf{r}} \hat{\mathbf{r}} \right)
+ (1 - ikr) \left( 3 \hat{\mathbf{r}} \hat{\mathbf{r}} - \mathbf{I} \right)
\right],
</math>
where <math>\hat{\mathbf{r}} = \mathbf{r} / |\mathbf{r}|</math> is the [[unit vector]] pointing from the source to the observation point.
This Green’s tensor describes the electric field generated by a dipole in a homogeneous medium. It is used to compute the off-diagonal blocks of the interaction matrix in DDA, that is, the interaction between distinct dipoles <math>j \ne k</math>. The singular self-term <math>\mathbf{G}(\mathbf{r} = 0)</math> is excluded and replaced by a prescribed local term involving the inverse polarizability tensor <math>\boldsymbol{\alpha}_j^{-1}</math>.
Thus, the electric field at dipole <math>j</math> due to dipole <math>k</math> is given by
:<math>
\mathbf{G}_{jk} = \frac{e^{ikr_{jk}}}{r_{jk}^3} \left[ k^2 r_{jk}^2 \left(\mathbf{I} - \hat{\mathbf{r}}_{jk} \hat{\mathbf{r}}_{jk}\right) + \left(1 - ikr_{jk}\right)\left(3\hat{\mathbf{r}}_{jk}\hat{\mathbf{r}}_{jk} - \mathbf{I}\right)\right],
</math>
where <math>\mathbf{r}_{jk} = \mathbf{r}_j - \mathbf{r}_k</math>, <math>r_{jk} = |\mathbf{r}_{jk}|</math>, and <math>\hat{\mathbf{r}}_{jk} = \mathbf{r}_{jk}/r_{jk}</math>. Here <math>\mathbf{I}</math> is the identity matrix and <math>k = 2\pi/\lambda</math> is the vacuum wavenumber.
Line 102 ⟶ 123:
{{Main|Polarizability}}
In the discrete dipole approximation, the electromagnetic response of a target is modeled by replacing the continuous material with a finite array of point dipoles. Each dipole represents a small volume of the material and acts as a polarizable unit that interacts with both the incident field and the fields radiated by all other dipoles. The key parameter that describes how each dipole responds to the local electric field is its polarizability <math>\alpha_j</math>. For a homogeneous material, the polarizability of a dipole is determined by the material’s complex dielectric function <math>\varepsilon(\lambda)</math>, which depends on the wavelength <math>\lambda</math> of light in vacuum. The dielectric function is related to the complex refractive index <math>n = n' + i n''</math> through <math>\varepsilon = n^2</math>. The goal in DDA is to assign to each dipole a polarizability <math>\alpha_j</math> such that the array of dipoles reproduces, as accurately as possible, the scattering and absorption behavior of the original continuous medium. For isotropic materials, a common starting point is the [[
In the discrete dipole approximation, the total volume of the target is divided into small cubic cells of volume <math>V_\mathrm{dipole} = d^3</math>, where <math>d</math> is the lattice spacing. The Clausius–Mossotti polarizability for each dipole is
:<math>
\
</math>
Line 120 ⟶ 141:
To improve the accuracy of the method various corrections to <math>\alpha_j</math> are applied. These include: the lattice dispersion relation (LDR) polarizability (Draine & Goodman, 1993), which adjusts <math>\alpha_j</math> to ensure that the dispersion relation of an infinite lattice of dipoles matches that of the continuous material; the radiative reaction (RR) correction, which compensates for the fact that each dipole radiates energy and is influenced by its own radiation field.
=== Size parameter ===
The size parameter is a dimensionless quantity used in scattering theory to characterize the size of a particle relative to the wavelength of the incident light. For a sphere, it is defined as:
:<math>x = \frac{2\pi r}{\lambda} = k r</math>
where: <math>x</math> is the size parameter (dimensionless), <math>a</math> is the radius of the sphere, <math>\lambda</math> is the wavelength of light in vacuum,
:<math>k = \frac{2\pi}{\lambda}</math> is the wavenumber.
In case of a sphere, the size parameter determines the scattering regime:
* If <math>x \ll 1</math>, [[Rayleigh scattering]] dominates.
* If <math>x \sim 1</math>, the scattering is in the regime of [[Mie scattering]].
* If <math>x \gg 1</math>, the [[Geometrical optics|geometric optics]] approximation becomes valid.
=== Effective radius and dipole discretization ===
For nonspherical targets with the same volume as a sphere, the effective radius <math>r_{\text{eff}}</math> is often used in place of <math>r</math>, with:
:<math>r_{\text{eff}} = \left( \frac{3 V_{\text{tot}}}{4\pi} \right)^{1/3} = \left( \frac{3 N d^3}{4\pi} \right)^{1/3}</math>
where: <math>N</math> is the total number of dipoles, <math>d</math> is the dipole spacing, <math>V_{\text{tot}} = N d^3</math> is the total volume represented by the dipoles. This gives effective size parameter
:<math>x_{\text{eff}} = \frac{2\pi r_{\text{eff}}}{\lambda}</math>
One convenient trick in certain DDA accuracy tests is to define wavelength as <math>2\pi</math>, in such a case effective radius is the same as effective size parameter.
=== Dipole-scale size parameter ===
Each polarizable point (dipole) occupies a cubic volume with side length <math>d</math>. Analogous to the global size parameter <math>x = 2\pi r / \lambda</math> used for whole particles, one can define a local size parameter for each dipole:
:<math>x_d = |m|kd = \frac{2\pi |m| d}{\lambda}</math>
This local parameter quantifies the ratio of the dipole size to the wavelength of light inside the material. For the DDA to be accurate, the field should vary slowly over the size of each dipole. This condition is satisfied when:
:<math>x_d = |m|kd \lesssim 0.5</math>
This ensures that each dipole is optically small, fields vary slowly over the dipole and the polarizability formula used for each dipole is accurate. Notice that a similar parameter plays a crucial role in the [[anomalous diffraction theory]] of van de Hulst, where the total phase shift experienced by light rays traveling through or around the particle is given by:
:<math>\delta = 2\pi (m - 1) \frac{r}{\lambda}</math>
This describes the optical path difference introduced by the particle (or in the case of DDA by a dipole).
=== Explicit Matrix Form of the DDA System ===
The Discrete Dipole Approximation (DDA) linear system is expressed as:
:<math> \mathbf{A} \mathbf{P} = \mathbf{E}_{\mathrm{inc}} </math>
where <math>\mathbf{A} \in \mathbb{C}^{3N \times 3N}</math> is the system matrix, <math>\mathbf{P} \in \mathbb{C}^{3N}</math> is the unknown polarization vector, <math>\mathbf{E}_{\mathrm{inc}} \in \mathbb{C}^{3N}</math> is the incident electric field vector.
We have
:<math>\mathbf{A} = -\mathbf{G} + \mathrm{diag}(\boldsymbol{\alpha}_1^{-1}, \dots, \boldsymbol{\alpha}_N^{-1})</math>.
<math>\mathbf{G}</math> encodes interactions between dipoles via the Green’s tensor (non-local), and <math> \mathrm{diag}(\boldsymbol{\alpha}_1^{-1}, \dots, \boldsymbol{\alpha}_N^{-1})</math> is a block-diagonal matrix with each block <math>\boldsymbol{\alpha}_j^{-1} \in \mathbb{C}^{3 \times 3}</math>.
Let ''N'' be the number of dipoles. Each dipole has a polarization vector <math> \mathbf{P}_j \in \mathbb{C}^3 </math>. The total system is a matrix equation of size <math> 3N \times 3N </math>:
:<math>
\begin{bmatrix}
\boldsymbol{\alpha}_1^{-1} & -\mathbf{G}_{12} & -\mathbf{G}_{13} & \cdots & -\mathbf{G}_{1N} \\
-\mathbf{G}_{21} & \boldsymbol{\alpha}_2^{-1} & -\mathbf{G}_{23} & \cdots & -\mathbf{G}_{2N} \\
-\mathbf{G}_{31} & -\mathbf{G}_{32} & \boldsymbol{\alpha}_3^{-1} & \cdots & -\mathbf{G}_{3N} \\
\vdots & \vdots & \vdots & \ddots & \vdots \\
-\mathbf{G}_{N1} & -\mathbf{G}_{N2} & -\mathbf{G}_{N3} & \cdots & \boldsymbol{\alpha}_N^{-1}
\end{bmatrix}
\begin{bmatrix}
\mathbf{P}_1 \\
\mathbf{P}_2 \\
\vdots \\
\mathbf{P}_N
\end{bmatrix}
=
\begin{bmatrix}
\mathbf{E}_{\mathrm{inc},1} \\
\mathbf{E}_{\mathrm{inc},2} \\
\vdots \\
\mathbf{E}_{\mathrm{inc},N}
\end{bmatrix}
</math>
Each block <math> \mathbf{A}_{jk} </math> is a <math> 3 \times 3 </math> complex matrix, defined by:
:<math>
\mathbf{A}_{jk} =
\begin{cases}
\boldsymbol{\alpha}_j^{-1} \in \mathbb{C}^{3 \times 3}, & \text{if } j = k \\
-\mathbf{G}_{jk} \in \mathbb{C}^{3 \times 3}, & \text{if } j \ne k
\end{cases}
</math>
So <math>\mathbf{A} \in \mathbb{C}^{3N \times 3N}</math> is composed of <math>N \times N</math> blocks, each of size <math>3 \times 3</math>. <math> \boldsymbol{\alpha}_j^{-1} \in \mathbb{C}^{3 \times 3} </math> is the inverse polarizability tensor, <math> \mathbf{G}_{jk} \in \mathbb{C}^{3 \times 3} </math> is the dyadic Green’s tensor for interaction between dipoles <math> j </math> and <math> k </math>, <math> \mathbf{P}_j, \mathbf{E}_{\mathrm{inc},j} \in \mathbb{C}^3 </math> are the dipole polarization and incident electric field at dipole <math> j </math>, respectively.
Typically dipoles are arranged on a regular grid. This implies translational invariance:
:<math> \mathbf{G}_{jk} = \mathbf{G}(\mathbf{r}_j - \mathbf{r}_k) = \mathbf{G}_{|j-k|} </math>
Because <math> \mathbf{G}_{|j-k|} = \mathbf{G}_{|k-j|} </math>, the matrix <math> \mathbf{A} </math> is symmetric:
:<math> \mathbf{A}_{jk} = \mathbf{A}_{kj} \quad \text{for all } j, k </math>
Each dipole has three vector components (<math>x</math>, <math>y</math>, <math>z</math>), so we can rearrange the unknown vector <math>\mathbf{P}</math> by grouping all x-components together, then y-components, then z-components:
:<math>
\mathbf{P} =
\begin{bmatrix}
\mathbf{P}_x \\
\mathbf{P}_y \\
\mathbf{P}_z
\end{bmatrix}
\in \mathbb{C}^{3N}
\quad \text{where} \quad
\mathbf{P}_x =
\begin{bmatrix}
P_{1x} \\
P_{2x} \\
\vdots \\
P_{Nx}
\end{bmatrix}, \quad
\mathbf{P}_y =
\begin{bmatrix}
P_{1y} \\
P_{2y} \\
\vdots \\
P_{Ny}
\end{bmatrix}, \quad
\mathbf{P}_z =
\begin{bmatrix}
P_{1z} \\
P_{2z} \\
\vdots \\
P_{Nz}
\end{bmatrix}
</math>
Similarly, the incident field can be grouped as:
:<math>
\mathbf{E}_{\mathrm{inc}} =
\begin{bmatrix}
\mathbf{E}_x^{\mathrm{inc}} \\
\mathbf{E}_y^{\mathrm{inc}} \\
\mathbf{E}_z^{\mathrm{inc}}
\end{bmatrix}
\in \mathbb{C}^{3N}
</math>
Because the system is linear, we can equivalently rewrite it in block matrix form, that describe how the <math>\beta</math>-component of polarization affects the <math>\alpha</math>-component of the resulting field:
<math>
\begin{bmatrix}
\mathbf{A}_{xx} & \mathbf{A}_{xy} & \mathbf{A}_{xz} \\
\mathbf{A}_{yx} & \mathbf{A}_{yy} & \mathbf{A}_{yz} \\
\mathbf{A}_{zx} & \mathbf{A}_{zy} & \mathbf{A}_{zz}
\end{bmatrix}
\begin{bmatrix}
\mathbf{P}_x \\
\mathbf{P}_y \\
\mathbf{P}_z
\end{bmatrix}
=
\begin{bmatrix}
\mathbf{E}_x^{\mathrm{inc}} \\
\mathbf{E}_y^{\mathrm{inc}} \\
\mathbf{E}_z^{\mathrm{inc}}
\end{bmatrix}
</math>
The expanded form of the equations is:
<math>
\begin{aligned}
-\mathbf{G}_{xx} \mathbf{P}_x - \mathbf{G}_{xy} \mathbf{P}_y - \mathbf{G}_{xz} \mathbf{P}_z +
(\boldsymbol{\alpha}^{-1} \mathbf{P})_x &= \mathbf{E}_x^{\mathrm{inc}} \\
-\mathbf{G}_{yx} \mathbf{P}_x - \mathbf{G}_{yy} \mathbf{P}_y - \mathbf{G}_{yz} \mathbf{P}_z +
(\boldsymbol{\alpha}^{-1} \mathbf{P})_y &= \mathbf{E}_y^{\mathrm{inc}} \\
-\mathbf{G}_{zx} \mathbf{P}_x - \mathbf{G}_{zy} \mathbf{P}_y - \mathbf{G}_{zz} \mathbf{P}_z +
(\boldsymbol{\alpha}^{-1} \mathbf{P})_z &= \mathbf{E}_z^{\mathrm{inc}}
\end{aligned}
</math>
Each block <math> \mathbf{G}_{ij} \in \mathbb{C}^{N \times N} </math> and the total system size is <math> 3N \times 3N </math>. The interaction matrix <math> \mathbf{G} \in \mathbb{C}^{3N \times 3N} </math> is composed of 9 blocks:
<math> \mathbf{G}_{xx}, \mathbf{G}_{xy}, \mathbf{G}_{xz}, \mathbf{G}_{yx}, \mathbf{G}_{yy}, \mathbf{G}_{yz}, \mathbf{G}_{zx}, \mathbf{G}_{zy}, \mathbf{G}_{zz} </math> (only 6 of them need to be evaluated due to symmetry).
Each matrix-vector multiplication <math>\mathbf{G}_{\alpha\beta} \mathbf{P}_\beta</math> can be computed as a convolution when the dipoles are arranged on a regular grid, allowing the use of Fast Fourier Transforms (FFTs) to accelerate the solution.
Let <math>\boldsymbol{\beta}_j = \boldsymbol{\alpha}_j^{-1}</math> denote the inverse polarizability tensor for dipole <math>j</math>. Each <math>\boldsymbol{\beta}_j</math> is a complex-valued <math>3 \times 3</math> matrix. This gives:
<math>
(\boldsymbol{\beta} \mathbf{P})_x =
\mathrm{diag}(\beta_{1,xx}, \dots, \beta_{N,xx}) \, \mathbf{P}_x +
\mathrm{diag}(\beta_{1,xy}, \dots, \beta_{N,xy}) \, \mathbf{P}_y +
\mathrm{diag}(\beta_{1,xz}, \dots, \beta_{N,xz}) \, \mathbf{P}_z
</math>
<math>
(\boldsymbol{\beta} \mathbf{P})_y =
\mathrm{diag}(\beta_{1,yx}, \dots, \beta_{N,yx}) \, \mathbf{P}_x +
\mathrm{diag}(\beta_{1,yy}, \dots, \beta_{N,yy}) \, \mathbf{P}_y +
\mathrm{diag}(\beta_{1,yz}, \dots, \beta_{N,yz}) \, \mathbf{P}_z
</math>
<math>
(\boldsymbol{\beta} \mathbf{P})_z =
\mathrm{diag}(\beta_{1,zx}, \dots, \beta_{N,zx}) \, \mathbf{P}_x +
\mathrm{diag}(\beta_{1,zy}, \dots, \beta_{N,zy}) \, \mathbf{P}_y +
\mathrm{diag}(\beta_{1,zz}, \dots, \beta_{N,zz}) \, \mathbf{P}_z
</math>
In the special case of an isotropic and homogeneous particle, the polarizabilities <math>\boldsymbol{\alpha}_j</math> are identical for all dipoles and proportional to the identity matrix: <math>\boldsymbol{\alpha}_j = \alpha \, \mathbf{I}</math>. Then, the inverse becomes <math>\boldsymbol{\beta}_j = \alpha^{-1} \, \mathbf{I}</math>, all off-diagonal elements vanish, and the expressions reduce to a simple element-wise division:
<math>
(\boldsymbol{\beta} \mathbf{P}) = \frac{1}{\alpha} \, \mathbf{P}
</math>
Note on practical implementation. In Fortran and MATLAB, arrays such as <math>\mathbf{P}(n_x,n_y,n_z,3)</math> or <math>\mathbf{E}^{\mathrm{inc}}(n_x,n_y,n_z,3)</math> are stored in column-major order, where the first index varies fastest in memory (anti-lexicographic). This means that all x-components <math>\mathbf{P}_x = \mathbf{P}(:,:,:,1)</math> are contiguous in memory, followed by all y-components <math>\mathbf{P}_y = \mathbf{P}(:,:,:,2)</math>, and then all z-components <math>\mathbf{P}_z = \mathbf{P}(:,:,:,3)</math>. In contrast, Python (NumPy) uses row-major order by default (lexicographic, last index varies fastest). To achieve the same contiguous layout of <math>\mathbf{P}_x</math>, <math>\mathbf{P}_y</math>, <math>\mathbf{P}_z</math> in memory, the array should be defined in Python as <math>\mathbf{P}(3, n_x, n_y, n_z)</math>, with the vector component index (x, y, z) first. This ensures that <math>\mathbf{P}_x = \mathbf{P}[0, :, :, :]</math> is stored contiguously in memory, followed by <math>\mathbf{P}_y</math> and <math>\mathbf{P}_z</math>.
== Green's function in Fourier space==
In the Fourier ___domain, the free-space dyadic Green's function for the vector Helmholtz equation can be written as
<math>
\widehat{\mathbf{G}}(\mathbf{k}) =
S(\mathbf{k})\left( \mathbf{I} - \frac{\mathbf{k}\,\mathbf{k}^{\mathrm{T}}}{|\mathbf{k}|^{2}} \right),
\qquad
S(\mathbf{k}) = \frac{1}{|\mathbf{k}|^{2} - k_{0}^{2} + i0}\,.
</math>
Here <math>\widehat{\mathbf{G}}(\mathbf{k})</math> is the Fourier transform of the dyadic Green's function <math>\mathbf{G}(\mathbf{r})</math>, <math>\mathbf{k} = (k_{x},k_{y},k_{z})</math> is the spectral wavevector, and <math>\mathbf{I}</math> is the identity tensor. The term <math>\mathbf{k}\,\mathbf{k}^{\mathrm{T}}/|\mathbf{k}|^{2}</math> is the longitudinal projector onto the direction of <math>\mathbf{k}</math>, so <math>\mathbf{I} - \mathbf{k}\,\mathbf{k}^{\mathrm{T}}/|\mathbf{k}|^{2}</math> is the transverse projector. The scalar factor <math>S(\mathbf{k})</math> contains the [[Helmholtz equation|Helmholtz denominator]] <math>|\mathbf{k}|^{2} - k_{0}^{2}</math>, where <math>k_{0} = \omega / c</math> is the wavenumber in the embedding medium. The infinitesimal imaginary term <math>+i0</math> implements the [[Sommerfeld radiation condition]], ensuring that the inverse Fourier transform produces an outgoing-wave solution.
==Conjugate gradient iteration schemes and preconditioning==
The solution of the linear system <math>\mathbf{A} \cdot \mathbf{P} = \mathbf{E}^{\mathrm{inc}}</math> in the DDA is typically performed using iterative methods. These methods aim to minimize the residual vector <math>\mathbf{r} = \mathbf{E}^{\mathrm{inc}} - \mathbf{A} \cdot \mathbf{P}</math> through successive approximations of the polarization vector <math>\mathbf{P}</math>. Among the earliest implementations were those based on direct matrix inversion,<ref name="purcell1973"/>
In practice, the dominant computational cost in DDA arises from the repeated evaluation of matrix-vector products during the iteration process. When the vector <math>\mathbf{P}</math> is stored in component-block form (as <math>\mathbf{P}_x</math>, <math>\mathbf{P}_y</math>, <math>\mathbf{P}_z</math>), the action of <math>\mathbf{A}</math> reduces to evaluating nine sub-products of the form <math>\mathbf{A}_{\alpha\beta} \mathbf{P}_\beta</math>, where <math>\alpha,\beta \in \{x,y,z\}</math>. These operations can be computed efficiently using convolution and FFT-based techniques when the dipole geometry is grid-based.
==Fast Fourier Transform for fast convolution calculations==
The use of the fast Fourier transform (FFT) to accelerate convolution operations in the discrete dipole approximation (DDA) was introduced by Goodman, Draine, and Flatau in 1991{{r|Goodman1991}}. Their method employed the three-dimensional FFT algorithm (GPFA) developed by Clive Temperton{{r|Temperton1983}}and required extending the interaction matrix from its original size <math>(n_x,n_y,n_z)</math> to <math>(2n_x,2n_y,2n_z)</math>. This extension was achieved by reversing and mirroring the Green’s function tensor blocks so that all positive and negative spatial offsets (lags) were represented in a single array, with an inserted zero plane between the positive and negative sides along each axis. This arrangement ensured that the discrete convolution of the Green’s function with the polarization vector could be performed as a cyclic convolution using FFTs, avoiding aliasing from wraparound effects. The sign-flipping of the Green’s function in the frequency ___domain and the block extension procedure became standard steps in efficient DDA implementations. Several alternative formulations have since been proposed.
A similar variant to that of Goodman, Draine, and Flatau was adopted in the 2021 MATLAB implementation by Shabaninezhad and Ramakrishna{{r|matlab2021}}. In this approach, the computational ___domain for the polarization vector is zero-padded to <math>(2n_x-1)\times (2n_y-1)\times (2n_z-1)</math> instead of the <math>2n_x\times 2n_y\times 2n_z</math> sizes used in ''DDSCAT''. The stored interaction matrix <math>G</math> differs from ''DDSCAT'' in that there is no zero plane inserted between the positive and negative offsets along each axis. The FFT is performed as a sequence of one-dimensional transforms along the <math>x</math>, <math>y</math>, and <math>z</math> axes, which is mathematically equivalent to performing a full 3-D FFT on the padded ___domain.
Sequence of 1D FFTs was used by MacDonald in his Ph. D. thesis <ref name=mcdonald2009/>.
Barrowes method is a numerical technique for multiplying an <math>n</math>-dimensional block Toeplitz matrix by a vector using the fast Fourier transform (FFT). In three dimensions with grid sizes <math>n_x</math>, <math>n_y</math>, and <math>n_z</math>, the method embeds the block Toeplitz array into a larger block circulant array of size <math>(2n_x-1)\times (2n_y-1)\times (2n_z-1)</math>, which ensures that the corresponding convolution is free of cyclic wraparound. The kernel—containing all positive and negative offsets with the self term set to zero—is reversed along the offset axes, flattened into a one-dimensional array, and transformed by a single long FFT. The input vector is likewise placed in a zero-padded ___domain of the same size, flattened, and transformed. An element-wise product in the frequency ___domain corresponds to the spatial-___domain convolution; an inverse FFT is then reshaped and cropped back to the physical ___domain to obtain the result. The method applies to arbitrary dimension <math>n</math> and block size, and was used originally in the discrete dipole approximation.<ref name="Barrowes2001" />
==Thermal discrete dipole approximation==
Line 148 ⟶ 417:
==References==
{{reflist|refs=
<ref name=chaumet2003>{{cite journal |last1=Chaumet |first1=Patrick C. |last2=Rahmani |first2=Adel |last3=Bryant |first3=Garnett W. |title=Generalization of the coupled dipole method to periodic structures |journal=Phys. Rev. B |volume=67 |issue=16 |pages=165404 |date=2003-04-02 |doi=10.1103/PhysRevB.67.165404 |arxiv=physics/0305051 |bibcode=2003PhRvB..67p5404C}}</ref>
Line 159 ⟶ 424:
<ref name=devoe1964>{{cite journal |last=DeVoe |first=Howard |title=Optical properties of molecular aggregates. I. Classical model of electronic absorption and refraction |journal=J. Chem. Phys. |volume=41 |issue=2 |pages=393–400 |date=1964-07-15 |doi=10.1063/1.1725879 |bibcode=1964JChPh..41..393D}}</ref>
<ref name=edalatpour2015>{{cite journal |last1=Edalatpour |first1=S. |last2=Čuma |first2=M. |last3=Trueax |first3=T. |last4=Backman |first4=R. |last5=Francoeur |first5=M. |title=Convergence analysis of the thermal discrete dipole approximation |journal=Phys. Rev. E |volume=91 |issue=6 |pages=063307 |year=2015 |doi=10.1103/PhysRevE.91.063307 |pmid=26172822 |arxiv=1502.02186 |bibcode=2015PhRvE..91f3307E}}</ref>
Line 184 ⟶ 448:
<ref name=yurkin2015>{{cite journal |last1=Yurkin |first1=M. A. |last2=Huntemann |first2=M. |title=Rigorous and fast discrete dipole approximation for particles near a plane interface |journal=J. Phys. Chem. C |volume=119 |issue=52 |pages=29088–29094 |year=2015 |doi=10.1021/acs.jpcc.5b09271 |url=https://scattering.ru/papers/Yurkin%20and%20Huntemann%20-%202015%20-%20Rigorous%20and%20fast%20discrete%20dipole%20approximation%20fo.pdf}}</ref>
<ref name=matlab2021>{{cite journal |last1=Shabaninezhad |first1=M. |last2=Awan |first2=M. G. |last3=Ramakrishna |first3=G. |title=MATLAB package for discrete dipole approximation by graphics processing unit: Fast Fourier Transform and Biconjugate Gradient |journal=J. Quant. Spectrosc. Radiat. Transfer |volume=262 |article-number=107501 |year=2021 |doi=10.1016/j.jqsrt.2020.107501 |bibcode=2021JQSRT.26207501S}}</ref>
Line 192 ⟶ 454:
<ref name=chaumet2024>{{cite journal |last=Chaumet |first=Patrick C. |title=A comparative study of efficient iterative solvers for the discrete dipole approximation |journal=J. Quant. Spectrosc. Radiat. Transfer |volume=312 |article-number=108816 |year=2024 |doi=10.1016/j.jqsrt.2023.108816 |bibcode=2024JQSRT.31208816C}}</ref>
<ref name=chaumet2022discrete>{{cite journal |last=Chaumet |first=Patrick C. |title=The discrete dipole approximation: A review |journal=Mathematics |volume=10 |issue=17 |pages=3049 |year=2022 |doi=10.3390/math10173049 |doi-access=free }}</ref>
<ref name=Barrowes2001>{{cite journal |last1=Barrowes |first1=B. E. |last2=Teixeira |first2=F. L. |last3=Kong |first3=J. A. |title=Fast algorithm for matrix–vector multiply of asymmetric multilevel block-Toeplitz matrices in 3-D scattering |journal=Microw. Opt. Technol. Lett. |volume=31 |issue=1 |pages=28–32 |year=2001 |doi=10.1002/mop.1348}}</ref>
<ref name=Goodman1991>{{cite journal |last1=Goodman |first1=J. J. |last2=Draine |first2=B. T. |last3=Flatau |first3=P. J. |title=Application of fast-Fourier-transform techniques to the discrete-dipole approximation |journal=Opt. Lett. |volume=16 |issue=15 |pages=1198–1200 |year=1991 |doi=10.1364/OL.16.001198 |pmid=19776919 |bibcode=1991OptL...16.1198G}}</ref>
<ref name="Petravic1979">{{cite journal |last1=Petravic |first1=M. |last2=Kuo-Petravic |first2=G. |title=An ILUCG algorithm which minimizes in the euclidean norm |journal=Journal of Computational Physics |volume=32 |issue=2 |pages=263–269 |year=1979|doi=10.1016/0021-9991(79)90133-5 |bibcode=1979JCoPh..32..263P |url=https://digital.library.unt.edu/ark:/67531/metadc1195635/ }}</ref>
<ref name="Moncada-Villa2022">{{Cite journal | last1 = Moncada-Villa | first1 = E. | last2 = Cuevas | first2 = J. C. | year = 2022 | title = Thermal discrete dipole approximation for near-field radiative heat transfer in many-body systems with arbitrary nonreciprocal bodies | journal = Physical Review B | volume = 106 | issue = 23 | pages = 235430 | doi = 10.1103/PhysRevB.106.235430 | arxiv = 2206.14921 | bibcode = 2022PhRvB.106w5430M }}</ref>
<ref name="Temperton1983">C. Temperton. "Self-sorting mixed-radix fast Fourier transforms." Journal of Computational Physics, 52.1 (1983): 1–23.</ref>
}}
| |||