Templates
kcorrect requires SED templates, which it fits to the given
photometry. There is a default set of templates which is limited and
works well at low redshifts. However, users can provide their own
templates. We show an example using a very broad set of templates that
is useful for observations that extend to the infrared.
If you would like to create custom templates, and have questions about the data format for the template files, please contact me!
Default Templates
The default set of templates are the five templates derived in the
kcorrect paper.
These were developed for kcorrect version v4.
They were designed to explain a variety of low redshift photometric data sets and some spectroscopic indices from the SDSS, and are only appropriate in the rest-frame ultraviolet through the near-infrared.
If you would like to examine these templates, the following code will
read them into a Template
object. This contains restframe_wave (in Angstroms) and
restframe_flux attributes. The flux is in erg/cm^2/s/A at 10 pc.
The corresponding values for the integrated star-formation history,
the remaining stellar mass, and other quantities, are available in the
object (and are read from the FITS file referenced below).
import kcorrect
import kcorrect.template
filename = os.path.join(kcorrect.KCORRECT_DIR, 'data',
'templates',
'kcorrect-default-v4.fits')
templates = kcorrect.template.Template(filename=filename)
Broad Set of Templates
For the purposes of examining mid-infrared active galactic nuclei, in Pai and Blanton (2024) we created a very large template set, including 57,600 simple stellar populations and 36 mid-IR active galactic nuclei models from CLUMPY, described in Nenkova et al (2008).
Although this sounds a bit ridiculously large, for reasonably sized samples it is tractable these days to fit non-negatively to large numbers of templates. There are two advantages of doing so. First, we do not have to reduce the dimensionality of the problem in any clever way; if there is a reasonable stellar population and active galactic nucleus fit to some set of photometry, the optimization will find it. Second, by using a Monte Carlo technique to estimate the resulting errors in the fits, we can obtain errors that more accurately reflect our lack of knowledge of the underlying variations in the templates. The disadvantage is of course that, though tractable, it is still pretty slow to do the fits.
The templates can be found at this Dropbox link (1.3 GB). If you have questions about the specific format of this file, please contact me. The code below demonstrates how to use this file. Note that these examples take some time to complete (of order an hour).
The first set of code pre-processes the templates in the context of a
particular redshift range of interest and a particular set of response
curves. This is an expensive operation, but it only need be performed
once. The template fitting calculation will involve an interpolation
of the resulting table over redshift, so you do not want the number of
redshifts (nredshift) to be too small. The Kcorrect object has a method tofits that allows you to save the
result to disk for later use.
import kcorrect.template
import kcorrect.kcorrect
responses = ['galex_FUV', 'galex_NUV',
'decam_g', 'decam_r', 'decam_z',
'wise_w1', 'wise_w2', 'wise_w3', 'wise_w4']
templates = kcorrect.template.Template(filename='templates_broad.fits')
kc = kcorrect.kcorrect.Kcorrect(responses=responses,
templates=templates,
redshift_range=[-0.002, 0.4],
nredshift=50)
kc.tofits('kcorrect_broad.fits')
This preprocessing step can take a lot of memory (up to about 20 GB),
because the both the templates and an interpolation object for all the
templates are stored. We can tell the Template class to not store
this object by specifying interpolate=False. In this case, the
interpolation necessary is done when the Kcorrect object is instantiated, and a full
interpolation object is never created. This is a bit slower
performance but uses much less memory (less than 4 GB).
import kcorrect.template
import kcorrect.kcorrect
responses = ['galex_FUV', 'galex_NUV',
'decam_g', 'decam_r', 'decam_z',
'wise_w1', 'wise_w2', 'wise_w3', 'wise_w4']
templates = kcorrect.template.Template(filename='templates_broad.fits',
interpolate=False)
kc = kcorrect.kcorrect.Kcorrect(responses=responses,
templates=templates,
redshift_range=[-0.002, 0.4],
nredshift=50)
kc.tofits('kcorrect_broad.fits')
Once the kcorrect_broad.fits file exists, it can be used to
actually perform template fitting and K-correction determination. Note
that whenever you are using a new set of responses or templates, you
should always make sure that the SED-fitting is adequately
reconstructing the original set of maggies you gave it; as shown
below, you can do that with the reconstruct method.
import kcorrect.kcorrect
redshift = 0.030317
maggies = [5.8345693e-09, 2.3105990e-08,
1.9847762e-06, 4.2561787e-06, 5.6498902e-06,
5.0805811e-06, 2.7973852e-06, 1.9943982e-06, 1.2780572e-06]
ivar = [1.11281764e+18, 1.61874720e+18,
2.82055401e+14, 6.13363357e+13, 3.48078746e+13,
4.30457938e+13, 1.41988431e+14, 1.22633740e+14, 4.96459265e+12]
kc = kcorrect.kcorrect.Kcorrect(filename='kcorrect_broad.fits')
# For this case, coeffs is large! [1, 57636]
# If you look carefully at this case, only 6 of the coefficients are non-zero!
coeffs = kc.fit_coeffs(redshift=redshift, maggies=maggies, ivar=ivar)
# Check the reconstructed maggies against the original
rmaggies = kc.reconstruct(redshift=redshift, coeffs=coeffs)
# We can then calculate the absolute magnitudes as usual
absmag = kc.absmag(redshift=redshift, maggies=maggies, ivar=ivar, coeffs=coeffs)