psphot Extended Source Fit Testing

This page discusses some experiments where we ran psphot on simulated images to test recent improvements to the psphot extended source fitting.

The 40 simulated images were provided last year by Peter Draper for use in testing the psphot kron radius and magnitude code. They were constructed using E Bertin's SkyMaker program.

The objects of interest are simulated galaxies which are described by the following parameters

  - x y   centre of object in pixels
  - mag   total magnitude
  - bt    build/disk ratio
  - br    bulge effective radii (half light arcsec)
  - bi    bulge inclination (always 1 in this sample)
  - ba    bulge position angle
  - dl    disk scale length (arcsec)
  - di    disk inclincation (always 1 in this sample)
  - da    disk position angle
  - z     redshift
  - ht    hubble type  (10 for pure disks, -6 for pure bulges)

These images were run though psphot using the STACKPHOT recipe. The parameters for the set of sources psphot selected for extended source fitting were matched by coordinates to the list of simulated galaxies. In the following sections we examine the fits for the 3 different models. We concentrate mostly on the simple galaxy models, pure disks (exponential profile) and and pure bulges (de Vaucouleur 1/4 profile).

Sersic Model

    I(r) = I(0) exp[ -k * R ** (1 / m) ]

psphot extended source parameters measured for PS_MODEL_SERSIC
   EXT_INST_MAG
   EXT_WIDTH_MAJ
   EXT_WIDTH_MIN
   EXT_THETA
   EXT_PAR_07       (0.5 / m)

Since all of the simulated galaxies in this sample have bulge inclination == 1 and disk inclination == 1, the major an minor axes should be equal and the angle will unmeasurable due to noise.

Sersic Index

If we look at the sersic index for all galaxies we get the following histogram. The hisogram on the right shows the counts for pure disks and bulges. We see that the fits for bulges are clustered around 4 as expected while disks mostly have values near 1.

Sersic Fit Magnitudes

The difference between the simulated galaxys' magnitudes and the sersic fit magnitudes are shown in the following plot. 91% of the sources are contained in the band near zero. The reason for the other bands needs to be investigated.

If we consider only the pure bulge sources the fits are better. psphot under estimates the flux somewhat. If we look at pure disks we see the banding.

The bands away from dmag == 0 are primarily objects with measured sersic index > 4.5 and small effective disk length.

Radii

disks

The models for disk galaxies are parametrized by disk scale length. The psphot extended source model width is scaled to the half light radius. To convert from EXT_WIDTH_MAJ to scale length we must divide by 1.7.

Here is a comparison of the measured width to model scale length for pure disk galaxies.

Plots of of the differences versus scale length and magnitude. ddl = dl/0.26 - EXT_WIDTH_MAJ/1.7

and a histogram of the difference in scale length (pixels)

bulges

If we compare the measured width to the bulge effective radius we find that the measured widths are a fraction smaller than expected.

Here is a histogram of the difference between br/0.26 and EXT_WIDTH_MAJ

Exponential Model

{{ Functional Form

I(r) = I(0) exp[ -k * R ]

psphot extended source parameters measured for PS_MODEL_EXP

EXT_INST_MAG EXT_WIDTH_MAJ EXT_WIDTH_MIN EXT_THETA

}}}

Since all of the simulated galaxies in this sample have bulge inclination == 1 and disk inclination == 1, the major an minor axes should be equal and the angle will unmeasurable due to noise.

Exponential magnitudes

In general the magnitude fit with exponential model is slightly less than the expected value. However when the galaxy sample is restricted to pure disks the magnitude differences are clustered around zero.

Radii

As with the sersice models, the measured width is about 2 x the modeled disk scale length.

de Vaucouleurs Model

The de Vaucouleurs model (DEV) provides a good fit to elliptical galaxies. In our sample these are represented by the galaxies with ht == -6 pure disks

    I(r) = I(0) exp[ -k * R ** (1 / 4) ]

psphot extended source parameters measured for PS_MODE_DEV
   EXT_INST_MAG
   EXT_WIDTH_MAJ
   EXT_WIDTH_MIN
   EXT_THETA

It looks like there is something wrong with the DEV models.The magnitude differences are all over the place and there was a problem matching up the models with the pure bulge objects.

modified DEV model

We noticed that the source position for the DEV models was different than those for the sersic or exponential so we changed the code to hold the source position fixed early in the fitting process GENE provide detail

In this experiment 3439 sources were matched, while in the previous only 260 were matched.