Imfit 2D-decomposition into structural components of PS1 candidate (dr1) images

Examples

The catalog contains 2D-decomposition of PS1 dr1 galaxies using Imfit.

Team

Alexandra Antipova (SAO RAS)
Dmitry Makarov (SAO RAS)
Lidia Makarova (SAO RAS)
Alexander Marchuk (SPbU)
Alexander Mosenkov (Pulkovo Observatory)
Vladimir Reshetnikov (SPbU)
Sergey Savchenko (SPbU)

Continuation of the work, the second version, is located at this link.

Classification of the 2D decomposition problems

We fulfilled revision of the problems with the Imfit 2D decomposition of our sample of the edge-on galaxies. The statistics is availabe by this link.

Visual inspection
Problem	#
Additional component is needed	841
Bulge is too big	622
Disk does not fit the profile	567
Asymmetric galaxy profile	544
Central part mimics a bulge	496
Bulge does not fit the profile	484
Central part is too flat	417
Model does not fit the profile	338
Bulge is redundant	300
Additional break is needed	223
Bulge is too flat and wide	189
Additional mask is needed	176
Bulge is missing	170
Spiral structure distorts the profile	170
Problem with the galaxy	84
Disk model is too complicated	50

Automatic inspection
Problem	#
scale equals error scale	3249
too small scale	2532
sersic scale greater than exponential	939
dsik flux << sersic flux	906
r_break2 < r_break1	216
Exponential disk PA error > 10	7

Gallery

It is a list of different difficulties with the 2D-decomposition.

Verification of the 2D-decomposition

Visual verification of the 2D-decomposition have been carried out. The statistics is availabe by this link. It is interesting to note that, despite some subjectivity in visual examination, the proportion of problematic profiles noted by different people was approximately the same for all examiner. 3322 galaxies (approximately 23%) sowed decomposition problems of different kind.

First run of the 2D-decomposition

We made first appempt of the 2D-decomposition of our saple of galaxies. During this stage an automatic pipeline used the results of the 1-D decomposition (bulge-to-total ratios, exponential scales of disks etc.) as an initial guess to perform the 2-D decomposition. The initial model to fit for each galaxy was the same model that was used during the 1-D decomposition, for example, if during the 1-D decomposition a model with a broken exponential disk was considered as the best choice for some object, the same broken exponential model (but in 2-D form) was used for that galaxy on this stage. When this initial fit is finished, the pipeline tries to modify the initial model in an attempt to find a better approximation for the galaxy. For example, if the initial model had no bulge, a small central component could be added to check for the presence of a faint bulge missed during the preliminary 1-D run. Another modification is the number of breaks in the exponential scale of the disk. If the model of the initial disk had no breaks, the pipeline attempts to add one, but if there are one or two breaks in the initial disk model, the pipeline tries to reduce their number to check if the simpler model could describe the disk. Therefore, for each galaxy we obtained several different models. After that, the pipeline performs the comparison between these models, using the BIC value, and some conditions on the relation between the derived model parameters (e.g. the upper limit on the bulge-to-total ratio, position of the disk breaks relative to the bulge dominated region etc.). This comparison yields us the best model for each galaxy.

Visual inspection of the 1D profile decomposition

We carried out a visual inspection of the profiles and 6067 of the 21240 profiles were marked as problematic. Examples of the problems are gathered in a short manual. The statistics on the inspection can be found here.

1D profile decomposition

An automatic decomposition of one-dimensional profiles has been carried out. The profile is a thin slice along the major axis of the galaxy folded with respect to the galaxy center. We use a set of different models: the simple exponential disk, the Sersic profile, the segmented function with two or three linear pieces, and their combinations. The best model was selected based on the Bayesian information criterion (BIC).