Mode Decomposition

Meep contains a feature to decompose arbitrary fields into a superposition of the harmonic modes of a given structure via its integration with the eigenmode solver MPB. This section provides an overview of the theory and implementation of this feature. Tutorial examples are provided in Tutorial/Mode Decomposition.

Theoretical Background

The theory underlying mode decomposition is described in Chapter 31 ("Modal methods for Maxwell's equations") of Optical Waveguide Theory by Snyder and Love.

Consider a waveguide with propagation axis along the direction and constant cross section in the transverse direction . For a given angular frequency ω we can solve for the eigenmodes of the structure. Thus, arbitrary fields of the form and can be decomposed into a basis of these eigenmodes:

β are the propagation wavevectors and α are the basis coefficients. Mode decomposition involves solving for these unknown quantities. The following steps are involved in the computation:

  1. In Meep, compute the Fourier-transformed fields and on a surface which is transverse to the waveguide and stored in a dft_flux object.

  2. In MPB, compute the eigenmodes and as well as the propagation wavevectors β for the same cross-sectional structure.

  3. Compute the coefficients α for any number of eigenmodes n=1,2,....

This is all done automatically in Meep using the get_eigenmode_coefficients routine.

Function Description

The mode-decomposition feature is available via the meep::fields::get_eigenmode_coefficients function callable from Python or C++. This function makes use of several lower-level functions which are described in more detail below. The C++ header for this function is:

void fields::get_eigenmode_coefficients(dft_flux flux,
                                        const volume &eig_vol,
                                        int *bands,
                                        int num_bands,
                                        int parity,
                                        double eig_resolution,
                                        double eigensolver_tol,
                                        std::complex<double> *coeffs,
                                        double *vgrp,
                                        kpoint_func user_kpoint_func,
                                        void *user_kpoint_data,
                                        vec *kdom_list,
                                        bool verbose)

The following are the parameters:

  • flux is a dft_flux object containing the frequency-domain fields on a cross-sectional slice perpendicular to the waveguide axis

  • eig_vol is the volume passed to MPB for the eigenmode calculation. In most cases this will simply be the volume over which the frequency-domain fields are tabulated (i.e. flux.where).

  • bands is an array of integers corresponding to the mode indices (equivalent to in the two formulas above)

  • num_bands is the length of the bands array

  • parity is the parity of the mode to calculate, assuming the structure has and/or mirror symmetry in the source region. If the structure has both and mirror symmetry, you can combine more than one of these, e.g. ODD_Z+EVEN_Y. This is especially useful in 2d simulations to restrict yourself to a desired polarization

  • eig_resolution is the spatial resolution to use in MPB for the eigenmode calculations

  • eigensolver_tol is the tolerance to use in the MPB eigensolver. MPB terminates when the eigenvalues stop changing by less than this fractional tolerance

  • coeffs is a user-allocated array of type std::complex<double> (shortened hereafter to cdouble) of length 2*num_freqs*num_bands where num_freqs is the number of frequencies stored in the flux object (equivalent to flux->Nfreq) and num_bands is the length of the bands input array. The expansion coefficients for the mode with frequency nf and band index nb are stored sequentially as α, α starting at slot 2*nb*num_freqs+nf of this array

  • vgrp is an optional user-allocated double array of length num_freqs*num_bands. On return, vgrp[nb*num_freqs + nf] is the group velocity of the mode with frequency nf and band index nb. If you do not need this information, simply pass NULL for this parameter.

  • user_kpoint_func is an optional function you supply to provide an initial guess of the wavevector of a mode with given frequency and band index having the following prototype:

  • verbose controls the verbosity of get_eigenmode. Defaults to false.

vec (*kpoint_func)(double freq, int mode, void *user_data);
  • user_kpoint_data is the user data passed to the user_kpoint_func

  • kdom_list is a user allocated array of meep::vec objects of length (num_bands * num_freqs). If non-null, this array is filled in with the wavevectors of the dominant planewave in the Fourier series expansion for each band from 1 to (num_bands * num_freqs). kdom_list[nb*num_freqs + nf] is the dominant planewave of the mode with frequency nf and band index nb. (Defaults to NULL.) This is especially useful for interpreting the modes computed in a uniform medium, because those modes are exactly planewaves proportional to where kdom is the wavevector.

 int num_bands = bands.size();
 int num_freqs = Flux->Nfreq;

 std::vector<cdouble> coeffs(2*num_bands*num_freqs);
 f.get_eigenmode_coefficients(...);

 for(int nb=0; nb<num_bands; nb++)
  for(int nf=0; nf<num_freqs++; nf++)
   { 
     // get coefficients of forward- and backward-traveling
     // waves in eigenmode bands[nb] at frequency #nf
     cdouble AlphaPlus = coeffs[2*nb*num_freqs+nf+0];
     cdouble AlphaMinus = coeffs[2*nb*num_freqs+nf+1];
     ...

Normalization

The α coefficients computed by get_eigenmode_coefficients are normalized such that their squared magnitude equals the power carried by the corresponding eigenmode:

where P is the power carried by the traveling eigenmode in the forward (+) or backward (-) direction. This is discussed in more detail below.

Besides get_eigenmode_coefficients, there are a few computational routines in libmeep that you may find useful for problems like those considered above.

Computing MPB Eigenmodes

  void *fields::get_eigenmode(double &omega,
                              direction d, const volume &where,
                              const volume &eig_vol,
                              int band_num,
                              const vec &kpoint, bool match_frequency,
                              int parity,
                              double resolution,
                              double eigensolver_tol,
                              bool verbose,
                              double *kdom);

Calls MPB to compute the band_numth eigenmode at frequency omega for the portion of your geometry lying in where which is typically a cross-sectional slice of a waveguide. kpoint is an initial starting guess for what the propagation vector of the waveguide mode will be. kdom, if non-NULL and length 3, is filled in with the dominant planewave for the current band (see above). This is implemented in mpb.cpp.

Working with MPB Eigenmodes

The return value of get_eigenmode is an opaque pointer to a data structure storing information about the computed eigenmode, which may be passed to the following routines:

// get a single component of the eigenmode field at a given point in space
std::complex<double> eigenmode_amplitude(const vec &p, void *vedata, component c);

// get the group velocity of the eigenmode 
double get_group_velocity(void *vedata);

// free all memory associated with the eigenmode
void destroy_eigenmode_data(void *vedata);

These functions are implemented in src/mpb.cpp.

Exporting Frequency-Domain Fields

  void output_dft(dft_flux flux, const char *HDF5FileName);

  void output_mode_fields(void *mode_data, dft_flux flux, const char *HDF5FileName);

output_dft exports the components of the frequency-domain fields stored in flux to an HDF5 file with the given filename In general, flux will store data for fields at multiple frequencies.

output_mode_fields is similar, but instead exports the components of the eigenmode described by mode_data which should be the return value of a call to get_eigenmode.

These functions are implemented in src/dft.cpp.

Computing Overlap Integrals

  std::complex<double> get_mode_flux_overlap(void *mode_data, 
                                             dft_flux *flux, 
                                             int num_freq, 
                                             std::complex<double>overlap[2]);

  std::complex<double> get_mode_mode_overlap(void *mode1_data,
                                             void *mode2_data,
                                             dft_flux *flux,
                                             std::complex<double>overlap[2]);

get_mode_flux_overlap computes the overlap integral between the eigenmode described by mode_data and the fields stored in flux for the num_freqth stored frequency, where num_freq ranges from 0 to flux->Nfreq-1. mode_data should be the return value of a previous call to get_eigenmode.

get_mode_mode_overlap is similar, but computes the overlap integral between two eigenmodes. mode1_data and mode2_data may be identical, in which case you get the inner product of the mode with itself. This should equal the group velocity of the mode based on the MPB's normalization convention.

These functions are implemented in src/dft.cpp.

How Mode Decomposition Works

The theoretical basis of the mode-decomposition algorithm is the orthogonality relation satisfied by the normal modes:

where the inner product involves an integration over transverse coordinates:

where is any surface transverse to the direction of propagation and is the unit normal vector to (i.e. in the case considered above). The normalization constant is a matter of convention, but in MPB it is taken to be the group velocity of the mode, , times the area of the cross-sectional surface : .

Now consider a Meep calculation in which we have accumulated frequency-domain fields and on a dft_flux object located on a cross-sectional surface . Invoking the eigenmode expansion and choosing the origin of the axis to be the position of the cross-sectional plane, the tangential components of the frequency-domain Meep fields take the form:

We have used the well-known relations between the tangential components of the forward- and backward-traveling field modes:

Taking the inner product of both equations with the and fields of each eigenmode, we find

Thus, by evaluating the integrals on the left-hand side of these equations — numerically, using the MPB-computed eigenmode fields and the Meep-computed fields as tabulated on the computational grid — and combining the results appropriately, we can extract the coefficients . This calculation is carried out by the routine fields::get_mode_flux_overlap. Although simple in principle, the implementation is complicated by the fact that, in multi-processor calculations, the Meep fields needed to evaluate the integrals are generally not all present on any one processor, but are instead distributed over multiple processors, requiring some interprocess communication to evaluate the full integral.

The Poynting flux carried by the Meep fields may be expressed in the form:

Thus, the power carried by a given forward- or backward-traveling eigenmode is given by:

or alternatively,

where . The are the eigenmode coefficients returned by get_eigenmode_coefficients.