What is PyAutoLens?#
PyAutoLens is open source software for the analysis and modeling of strong gravitational lenses, with its target audience anyone with an interest in astronomy and cosmology.
The software comes distributed with the HowToLens Jupyter notebook lectures, which are written assuming no previous knowledge about what gravitational lensing is and teach a new user theory and statistics required to analyse strong lens data. Checkout the howtolens section of the readthedocs.
An overview of PyAutoLens’s core features can be found in the overview section of the readthedocs.
Strong Gravitational Lensing#
When two galaxies are aligned down the line-of-sight to Earth, light rays from the background galaxy are deflected by the intervening mass of one or more foreground galaxies. Sometimes its light is fully deflected around the foreground galaxies, traversing multiple paths to the Earth, meaning that the background galaxy is observed multiple times. This alignment of galaxies is called a strong gravitational lens, an example of which, SLACS1430+4105, is shown in the image below. The massive elliptical lens galaxy can be seen in the centre of the left panel, surrounded by a multiply imaged source galaxy whose light has been distorted into an ‘Einstein ring’. The central and right panels shows reconstructions of the source’s lensed and unlensed light distributions, which are created using a model of the lens galaxy’s mass to trace backwards how the source’s light is gravitationally lensed.
Strong lensing provides astronomers with an invaluable tool to study a diverse range of topics, including the structure of galaxies, dark matter and the expansion of the Universe.
The past decade has seen the discovery of many hundreds of new strong lenses, however the modeling of a strong lens is historically a time-intensive process that requires significant human intervention to perform, restricting the scope of any scientific analysis. In the next decade of order of one hundred thousand strong lenses will be discovered by surveys such as Euclid, the Vera Rubin Observatory and Square Kilometer Array.
The goal of PyAutoLens is to enable fully automated strong lens analysis, such that these large samples of strong lenses can be exploited to their fullest.
How does PyAutoLens Work?#
A strong lens system can be quickly assembled from abstracted objects. A
Galaxy object contains one or
MassProfile’s, which represent its two dimensional distribution of starlight and mass.
Galaxy’s lie at a particular distance (redshift) from the observer, and are grouped into
Plane’s is achieved by passing them to a
Tracer with an
astropy Cosmology. By passing
these objects a
Grid2D strong lens sightlines are computed, including multi-plane ray-tracing. All of these
objects are extensible, making it straightforward to compose highly customized lensing system. The example code
below shows this in action:
import autolens as al import autolens.plot as aplt from astropy import cosmology as cosmo """ To describe the deflection of light by mass, two-dimensional grids of (y,x) Cartesian coordinates are used. """ grid = al.Grid2D.uniform( shape_native=(50, 50), pixel_scales=0.05, # <- Conversion from pixel units to arc-seconds. ) """ The lens galaxy has an elliptical isothermal mass profile and is at redshift 0.5. """ mass = al.mp.Isothermal( centre=(0.0, 0.0), ell_comps=(0.1, 0.05), einstein_radius=1.6 ) lens_galaxy = al.Galaxy(redshift=0.5, mass=mass) """ The source galaxy has an elliptical exponential light profile and is at redshift 1.0. """ disk = al.lp.Exponential( centre=(0.3, 0.2), ell_comps=(0.05, 0.25), intensity=0.05, effective_radius=0.5, ) source_galaxy = al.Galaxy(redshift=1.0, disk=disk) """ We create the strong lens using a Tracer, which uses the galaxies, their redshifts and an input cosmology to determine how light is deflected on its path to Earth. """ tracer = al.Tracer.from_galaxies( galaxies=[lens_galaxy, source_galaxy], cosmology: al.cosmo.Planck15() ) """ We can use the Grid2D and Tracer to perform many lensing calculations, for example plotting the image of the lensed source. """ tracer_plotter = aplt.TracerPlotter(tracer=tracer, grid=grid) tracer_plotter.figures_2d(image=True)
To perform lens modeling, PyAutoLens adopts the probabilistic programming
language PyAutoFit. PyAutoFit allows users to compose a
lens model from
Galaxy objects, customize the model parameterization and
fit it to data via a non-linear search (e.g. dynesty,
emcee or PySwarms). The example
code below shows how to setup and fit a lens model to a dataset:
import autofit as af import autolens as al import autolens.plot as aplt """ Load Imaging data of the strong lens from the dataset folder of the workspace. """ imaging = al.Imaging.from_fits( data_path="/path/to/dataset/image.fits", noise_map_path="/path/to/dataset/noise_map.fits", psf_path="/path/to/dataset/psf.fits", pixel_scales=0.1, ) """ Create a mask for the imaging data, which we setup as a 3.0" circle, and apply it. """ mask = al.Mask2D.circular( shape_native=imaging.shape_native, pixel_scales=imaging.pixel_scales, radius=3.0 ) imaging = imaging.apply_mask(mask=mask) """ We model the lens galaxy using an elliptical isothermal mass profile and the source galaxy using an elliptical sersic light profile. To setup these profiles as model components whose parameters are free & fitted for we set up each Galaxy as a `Model` and define the model as a `Collection` of all galaxies. """ # Lens: mass = af.Model(al.mp.Isothermal) lens = af.Model(al.Galaxy, redshift=0.5, mass=lens_mass_profile) # Source: disk = af.Model(al.lp.Sersic) source = af.Model(al.Galaxy, redshift=1.0, disk=disk) # Overall Lens Model: model = af.Collection(galaxies=af.Collection(lens=lens, source=source)) """ We define the non-linear search used to fit the model to the data (in this case, Dynesty). """ search = af.DynestyStatic(name="search[example]", nlive=50) """ We next set up the `Analysis`, which contains the `log likelihood function` that the non-linear search calls to fit the lens model to the data. """ analysis = al.AnalysisImaging(dataset=imaging) """ To perform the model-fit we pass the model and analysis to the search's fit method. This will output results (e.g., dynesty samples, model parameters, visualization) to hard-disk. """ result = search.fit(model=model, analysis=analysis) """ The results contain information on the fit, for example the maximum likelihood model from the Dynesty parameter space search. """ print(result.samples.max_log_likelihood())
To get started, users can check-out the PyAutoLens’s rich feature-set by going through the
of our readthedocs. This illustrates the API for all of PyAutoLens’s core features, including how to simulate
strong lens datasets, reconstructing the lensed source galaxy on adaptive pixel-grids and fitting interferometer
For new PyAutoLens users, we recommend they start by
installing PyAutoLens (if you haven’t
already!), read through the
introduction.ipynb notebook on
the autolens_workspace and take the
HowToLens Jupyter notebook lecture series on
strong gravitational lensingtick_maker.min_value.