bbf 0.2.0

Fast computation of broadband fluxes and mags

Broadband fluxes (bbf)

---

Table of contents

• Installation
• [Getting started](#getting started)
• License

A module to evaluate the broadband fluxes and magnitudes of spectrophotometric standards.

For the moment, bbf relies on a modified version of sncosmo for passbands and magsys definition.

Installation

virtual environments

We recommend using conda which comes with a compiled version of suitesparse``. venv` is also a suitable option if suitesparse is already installed on your machine, or if you are ready to compile it yourself.

As a reminder:

conda create -n MY_ENV
conda activate MY_ENV

or:

python -m venv MY_ENV
source MY_ENV/bin/activate

Prerequisites

conda packages for bbf are in preparation (but not ready yet). Better install them directly in conda.

conda install ipython numpy scipy matplotlib scikit-sparse pandas h5py pyarrow

Until our changes are merged in sncosmo, we use a modified version of it:

git clone https://github.com/nregnault/sncosmo.git
cd sncosmo
pip install . 

Installing bbf

pip install bbf

if you prefer installing from sources:

git clone clone [e-mail]:lemaitre/bbf.git
cd bbf
pip install -e . 

Installing the Lemaitre bandpasses

If you plan to use the latest version of the megacam6, ztf and hsc passbands, install the lemaitre.bandpasses package:

pip install lemaitre-bandpasses

or

git clone https://gitlab.in2p3.fr/lemaitre/lemaitre/bandpasses
cd bandpasses
git lfs pull
pip install . 

Getting started

The goal of bbf is to efficiently compute broadband fluxes and magnitudes, i.e. quantities of the form:

$$f_{xys} = \int S(\lambda) \lambda T_{xys}(\lambda) d\lambda$$

where $\lambda$ is the SED of an object, $T_{xyz}(\lambda)$ is the bandpass of the instrument used to observe it. $T$ may depend on the focal plane position of the object and, if the focal plane is a mosaic of sensors, on the specific sensor $s$ where the observation is made. In practice, $x,y$ are coordinates, in pixels, in the sensor frame, and $s$ is a unique sensor index (or amplifier index).

Computing magnitudes requires an additional ingredient: the flux of a reference spectrum $S_{ref}(\lambda)$, usually the AB spectrum, integrated in the same passband (same sensor, same position).

$$m = -2.5 \log_{10} \left(\frac{\int S(\lambda) \lambda T_{xyz}(\lambda) d\lambda}{\int S_{ref}(\lambda) \lambda T_{xyz}(\lambda) d\lambda}\right)$$

To compute these integrales, bbf uses the technique implemented in nacl, which consists in projecting the bandpasses and SED on spline bases:

$$S(\lambda) = \sum_i \theta_i {\cal B}_i(\lambda)$$

and

$$T(\lambda) = \sum_j t_j {\cal B}_j(\lambda)$$

If we precompute the products $G_{ij} = \int \lambda {\cal B}i(\lambda){\cal B}_j(\lambda) d\lambda$ the integrals above can be expressed as a simple contraction:

$$f = \theta_i G_{ij} t_j$$

where $G$ is very sparse, since the B-Splines ${\cal B}_i$ have a compact support. If the bandpass $T$ is spatially variable, the $t_j$ coefficients are themselves developped on a spatial spline basis.

$$t_j = \sum_{kj} \tau_{kj} {\cal K}(x,y)$$

The contraction above is then of the form:

FilterSets and StellarLibs

bbf implements two main kind of objects: FilterLib, which holds a set of band passes, projected on spline bases (${\cal K_j(x,y)}$ and ${\cal B}i(\lambda)$), and StellarLib which manages a set of spectra, also projected on a spline basis (not necessily the splines used for the filters).

Loading a filter lib

Building a complete version of a FilterLib requires some care. The standard FilterLib used in the Lemaître analysis is build and maintained within the package lemaitre.bandpasses. To access it:

from lemaitre import bandpasses

flib = bandpasses.get_filterlib()

The first time this function is called, the `FilterLib`` is built and cached. The subsequent calls access the cached version, and never take more than a few milliseconds.

Loading Stellar Libraries

As of today, bbf implements two kinds of StellarLibs: pickles and Calspec. An interface to gaiaXP is in development.

To load the pickles library:

import bbf.stellarlib.pickles
pickles = bbf.stellarlib.pickles.fetch()

To load the most recent version of Calspec:

import bbf.stellarlib.calspec
calspec = bbf.stellarlib.calspec.fetch()

Computing Broadband fluxes

With a FilterSet and a StellarLib in hand, one can compute broadband fluxes and broadband mags.

Broadband fluxes

import bbf.stellarlib.pickles
from lemaitre import bandpasses

flib = bandpasses.get_filterlib()
pickles = bbf.stellarlib.pickles.fetch()

# number of measurements
nmeas = 100_000

# which stars ? 
star = np.random.choice(np.arange(0, len(pickles)), size=nmeas)

# in which band ?
band = np.random.choice(['ztf::g', 'ztf::r', 'ztf::I'], size=nmeas)

# observation positions
x = np.random.uniform(0., 3072., size=nmeas)
y = np.random.uniform(0., 3080., size=nmeas)
sensor_id = np.random.choice(np.arange(1, 65), size=nmeas)

fluxes = flib.flux(pickles, star, band, x=x, y=y, sensor_id=sensor_id)

Broadband magnitudes

To convert broadband fluxes into broadband magnitudes, we need to compute the reference fluxes, in the same effective measurement band passes. This is done using an auxiliary object called MagSys:

from bbf.magsys import SpecMagSys 
import bbf.stellarlib.pickles
from lemaitre import bandpasses

flib = bandpasses.get_filterlib()
pickles = bbf.stellarlib.pickles.fetch()

# number of measurements
nmeas = 100_000

# which stars ? 
star = np.random.choice(np.arange(0, len(pickles)), size=nmeas)

# in which band ?
band = np.random.choice(['ztf::g', 'ztf::r', 'ztf::I'], size=nmeas)

# observation positions
x = np.random.uniform(0., 3072., size=nmeas)
y = np.random.uniform(0., 3080., size=nmeas)
sensor_id = np.random.choice(np.arange(1, 65), size=nmeas)

ms = SpecMagSys('AB')
mags = ms.mag(pickles, star, band, x=x, y=y, sensor_id=sensor_id)

License

bbf is distributed under the terms of the MIT license.