spectral_models module

Provides classes for modelling the 21-cm spectral line emitted by a SPH particle.

class martini.spectral_models.DiracDeltaSpectrum(ncpu: int | None = None, spec_dtype: type = <class 'numpy.float64'>)

Bases: _BaseSpectrum

Class implemeting a Dirac-delta model for the spectrum of the HI line.

The line is modelled as a Dirac-delta function, centered at the particle velocity.

Parameters:

• ncpu (int, optional) – Number of cpus to use for evaluation of particle spectra. Defaults to 1 if not provided.

• spec_dtype (type, optional) – Data type of the arrays storing spectra of each particle, can be used to manage memory usage by adjusting precision.

evaluate_spectra(source: SPHSource, datacube: DataCube, mask: slice | EllipsisType = Ellipsis) → Annotated[Quantity, Unit(dimensionless)]

Evaluate the spectra.

Separated into this function so that it can be called by a parallel process pool.

Parameters:

• source (SPHSource) – Source object containing arrays of particle properties.

• datacube (DataCube) – DataCube object defining the observational parameters, including spectral channels.

• mask (slice, optional) – Slice defining the subset of particles to operate on.

Returns:

The evaluated dimensionless spectra.

Return type:

Quantity

get_spectral_function_extra_data(source: SPHSource, datacube: DataCube, mask: slice | EllipsisType = Ellipsis, extra_data: dict[str, Quantity] | None = None) → dict[str, Quantity]

Initialize extra data needed by spectral function. Default is no extra data.

Derived classes should override this function, if needed, to populate the dict with any information from the source that is required by the spectral_function(), then call super().get_spectral_function_extra_data with the extra_data argument set to the dictionary that they loaded. This function then handles setting up the array for broadcasting.

Parameters:

• source (SPHSource) – Source object, making particle properties available.

• datacube (DataCube) – DataCube object defining the observational parameters, including spectral channels.

• mask (slice, optional) – Slice defining the subset of particles to operate on.

• extra_data (dict, optional) – dict containing additional data arrays needed for the spectral function evaluation. The dictionary entries should contain scalars or 1D arrays of particle properties.

Returns:

The extra data that have been read in and prepared for use.

Return type:

dict

See also

martini.spectral_models.GaussianSpectrum.get_spectral_function_extra_data

half_width(source: SPHSource) → Annotated[Quantity, Unit('km / s')]

Dirac-delta function has 0 width.

Parameters:

source (SPHSource) – Source object, making particle properties available.

Returns:

Quantity, with dimensions of velocity. Velocity dispersion of 0 * U.km * U.s**-1.

Return type:

Quantity

init_spectra(source: SPHSource, datacube: DataCube) → None

Pre-compute the spectrum of each particle.

The spectral model defined in spectral_function() is evaluated using the channel edges from the DataCube instance and the particle velocities of the SPHSource (or derived class) instance provided.

If the instance of this class was initialized with ncpu > 1 then a process pool is created to distribute subsets of the calculation in parallel. To minimize overhead form serializing large amounts of data in multiprocess communications, each parallel process inherits the entire line-of-sight velocity array (cheap because of copy-on-write behaviour), then masks its copy to the subset to operate on.

Parameters:

• source (SPHSource) – Source object containing arrays of particle properties.

• datacube (DataCube) – DataCube object defining the observational parameters, including spectral channels.

spectral_function(a: Annotated[Quantity, Unit('km / s')], b: Annotated[Quantity, Unit('km / s')], vmids: Annotated[Quantity, Unit('km / s')], extra_data: dict[str, Quantity] | None = None) → Annotated[Quantity, Unit(dimensionless)]

Evaluate a Dirac-delta function in a channel.

Parameters:

• a (Quantity) – Quantity, with dimensions of velocity. Lower spectral channel edge(s).

• b (Quantity) – Quantity, with dimensions of velocity. Upper spectral channel edge(s).

• vmids (Quantity) – Quantity, with dimensions of velocity. Particle velocities along the line of sight.

• extra_data (dict, optional) – dict containing arrays of extra data for the spectral function evaluation.

Returns:

The evaluated spectral model (dimensionless).

Return type:

Quantity

class martini.spectral_models.GaussianSpectrum(sigma: Quantity, Unit("km / s")]=<Quantity 7. km / s>, ncpu: int | None = None, spec_dtype: type = <class 'numpy.float64'>)

Bases: _BaseSpectrum

Class implementing a Gaussian model for the spectrum of the HI line.

The line is modelled as a Gaussian of either fixed width, or of width scaling with the particle temperature as \(\\sqrt{k_B T / m_p}\), centered at the particle velocity.

Parameters:

• sigma (Quantity or str, optional) – Quantity, with dimensions of velocity, or string "thermal". Width of the Gaussian modelling the line (constant for all particles), or specify "thermal" for width equal to \(\\sqrt{k_B T / m_p}\) where \(k_B\) is Boltzmann’s constant, \(T\) is the particle temperature and \(m_p\) is the particle mass.

• ncpu (int, optional) – Number of cpus to use for evaluation of particle spectra. Defaults to 1 if not provided.

• spec_dtype (type, optional) – Data type of the arrays storing spectra of each particle, can be used to manage memory usage by adjusting precision.

See also

martini.spectral_models._BaseSpectrum, martini.spectral_models.DiracDeltaSpectrum

evaluate_spectra(source: SPHSource, datacube: DataCube, mask: slice | EllipsisType = Ellipsis) → Annotated[Quantity, Unit(dimensionless)]

Evaluate the spectra.

Separated into this function so that it can be called by a parallel process pool.

Parameters:

• source (SPHSource) – Source object containing arrays of particle properties.

• datacube (DataCube) – DataCube object defining the observational parameters, including spectral channels.

• mask (slice, optional) – Slice defining the subset of particles to operate on.

Returns:

The evaluated dimensionless spectra.

Return type:

Quantity

get_spectral_function_extra_data(source: SPHSource, datacube: DataCube, mask: slice | EllipsisType = Ellipsis, extra_data: dict[str, Quantity] | None = None) → dict[str, Quantity]

Expose particle velocity dispersions.

Access to these is needed by spectral_function().

Parameters:

• source (SPHSource) – Source object.

• datacube (DataCube) – DataCube object.

• mask (slice, optional) – Slice defining the subset of particles to operate on.

• extra_data (dict, optional) – dict containing arrays of extra data for the spectral function evaluation.

Returns:

The extra data that have been read in and prepared for use.

Return type:

dict

half_width(source: SPHSource) → Annotated[Quantity, Unit('km / s')]

Get 1D velocity dispersions from particle temperatures, or return constant.

Parameters:

source (SPHSource) – Source object, making particle properties available.

Returns:

Quantity, with dimensions of velocity. Velocity dispersion (constant, or per particle).

Return type:

Quantity

init_spectra(source: SPHSource, datacube: DataCube) → None

Pre-compute the spectrum of each particle.

The spectral model defined in spectral_function() is evaluated using the channel edges from the DataCube instance and the particle velocities of the SPHSource (or derived class) instance provided.

If the instance of this class was initialized with ncpu > 1 then a process pool is created to distribute subsets of the calculation in parallel. To minimize overhead form serializing large amounts of data in multiprocess communications, each parallel process inherits the entire line-of-sight velocity array (cheap because of copy-on-write behaviour), then masks its copy to the subset to operate on.

Parameters:

• source (SPHSource) – Source object containing arrays of particle properties.

• datacube (DataCube) – DataCube object defining the observational parameters, including spectral channels.

spectral_function(a: Annotated[Quantity, Unit('km / s')], b: Annotated[Quantity, Unit('km / s')], vmids: Annotated[Quantity, Unit('km / s')], extra_data: dict[str, Quantity] | None = None) → Annotated[Quantity, Unit(dimensionless)]

Evaluate a Gaussian integral in a channel.

Requires sigma to be available from spectral_function_extra_data.

Parameters:

• a (Quantity) – Quantity, with dimensions of velocity. Lower spectral channel edge(s).

• b (Quantity) – Quantity, with dimensions of velocity. Upper spectral channel edge(s).

• vmids (Quantity) – Quantity, with dimensions of velocity. Particle velocities along the line of sight.

• extra_data (dict, optional) – dict containing arrays of extra data for the spectral function evaluation.

Returns:

The evaluated spectral model (dimensionless).

Return type:

Quantity