Hubble Source Catalog (HSC) Use Case # 4 - Photometry of Slightly Resolved Objects

Reference:- "The Color-Magnitude Relation for Metal-Poor Globular Clusters in M87"

- Peng et al. (2009) ApJ, 703, 42

1. How to select reasonable parameters for NumImages
2. Combining ACS and WFPC2 magnitudes
3. The effect of using MagAper2 or MagAuto
4. Techniques for making the HSC sample more uniform

Part 1: View the Data

• STEP 1 - Enter the Hubble Legacy Archive website and search for M87.
  
  In the inventory you will see properties of the individual images such as RA and Dec, Image ID, which filter the image was taken with etc. These columns can can filtered by entering key words just below the table headings.
  
  

Start refining the results by entering ACS/WFC in the Detector field for ACS Wide Field Camera. This should take your 1726 results down to 259. Now enter *F814W* in the spectral_elt field (spectral element), where including the "wildcards" displays all images that include the text F814W.

Go into the images tab and enter the interactive display of the first image, HST_10543_01_ACS_WFC_F814W.

Click on the HSC (beta) check box to see all the HSC sources on this image (may take a short while).





The HSC combines visit based source lists from all available WFPC2 and ACS/WFC data for a given field. Because of this however, the small possibility of having artifacts and occasional unmatched source readings becomes a very real possibility in the final combined image. To remove most of the false detections, enter 5 in the Require NumImages field, under the advanced HSC controls. Note that there are not many sources in the nucleus of M87. The current generation of ACS and WFPC2 SExtractor source lists have problems identifying soures in regions of high background. In addition, just outside of this region many false detections are present. See FAQ under the limitations section for further discussion of the effects of high background.


• STEP 2- Distinguish between point and extended sources.
  
  Clicking on a source brings up the HSC summary data. We can use the concentration index (the difference in magnitude of a source at a radius of 3 pixels to 1) to quantify the source's type of structure. The magnitude of the source for each filter is also found here (measured in ABMAG). The HSC gathers source photometry from a visit based method. As we will see later this imposes a limit on faint sources that can be detected relative to studies which combine images to form deep mosaics. Having data for each visit can be useful in cases such as studying the variability of sources.

  
  

  
  
  This image was constructed using additional software for illustrative purposes.
  
  Highlighted are some point sources (green) as well as extended sources (blue) accompanied by their associated CI values. CI normally spans a range of 0.8 to 2.5 mag, where objects with CI ≲ 1.1 mag are point sources, CI ≳ 1.1 mag are extended sources.
  
  Note that M87 is densely populated with GCs, there are only a few 10s of foreground stars and 1000s of GCs. A good technique to locate bright point sources by eye is to look for diffraction spikes like the star in the right most square section of the diagram or an Airy disk like the star in the top square.
  
  Note also that these CI values are averages of all entries in the HSC (i.e. include all instruments and filters). There is no attempt to match the size of the aperture in terms of arc seconds for example. Different instruments have different scales (e.g. ACS is 0.05"/pix while the WFC chips of the WFPC2 have a scale of 0.10"/pix). Hence the resulting values that go into this average CI are very non uniform. The more specific CI values can be found in the detailed form in step 3.
  
  
• STEP 3- Examine the data from the HSC.
  
  Open up the Summary Form located on the HSC homepage http://archive.stsci.edu/hst/hscv1/.
  
  Use the following criteria:
  
  Target Name = M87
  Radius = 5 (arcmin)
  Mag Type = MagAper2
  NumImages > 5
  Maximum records = 5001
  
  
  and hit search.
  
  The table displayed contains all the information from tens of visits for each source conveniently located in one table.
  
  The MagAuto SExtractor values are an attempt to determine the total magnitudes of objects. Since we are primarily interested in determining the V-I color of the clusters it is probably more accurate to use aperture magnitudes hence the choice of MagAper2 (using a 3 pixel radius). Later in this exercise we compare results using MagAper2 and MagAuto.
  
  The query should return 3065 results. The first source displayed is the nucleus, hence why there is a large number of images.

  

  
  
  Click the matchID of the first source and it will take you to the DETAILED FORM where information for each individual image is displayed.
  
  Note that in this example a large amount of the images were taken in experiment 10543. To view the abstract of this proposal, in the HLA reopen the inventory view for M87 and in the prop ID field filter the results to display only 10543. Follow the hyperlink on any of those 10543 proposal IDs.
  
  

  

  
  
  From analysis of this table it is clear that there is an abundance of ACS F814W (A_F814W) data but no ACS F606W (A_F606W). We will have to use WFPC2 F606W (W_F606W) data instead, unlike Peng et al. who exclusively used ACS. However the inventory view of the HLA shows 57 ACS F606W observations (to check this yourself simply search F606W in step 1 instead of F814W). The reason the A_F606W values are missing in the HSC is because a single image was taken in each visit. This makes it impossible to remove cosmic rays. However the A_F814W data used 4 exposures per visit, and so a high quality source list was made by the HLA. Which was used to determine the locations to measure photometry in the A_F606W image as well. The HSC software was not smart enough to know this and ignored the HLA A_F606W source list. This rule is being changed in the HSC and hence the A_606W data will be included in the next release.
  
  Click on "Display numeric columns graphically using VOPlot". It may take a while to load the data because of the large sample size. The display will say "No VOTable loaded" until the data appear. Set the x-axis to wavelength and the y-axis to MagAuto. We see several vertical lines of data for each of the filters. This gives us a feeling for what filters the data has been taken in. We find that a large number of observations have been taken in the F814W filter. We also note as above that the large numbers of A_F606W observations have not made it into the HSC.

  

  
  Now go back to SUMMARY FORM and use VOPlot to plot the color-magnitude diagram of M87. Open the functions tab in VOTable and select 'create new columns'. Since there is no A_F606W data we are forced to use magnitudes from WFPC2 which has less depth. Type $7 - $9 into the expression section. Enter a column name like 'V-I' and click add. Now plot V-I against W_F814W. To match the format of Peng et al. select the option to reverse the y-axis and change the limits with the 'format plot' button.

  

  
  
  Note that this plot was constructed using only the HSC and the software contained within. While this is often sufficient, you may want to use your own favorite software package (e.g. IDL, IRAF, TOPCAT, ... ) to do more sophisticated analysis. We can see some of the key features in Peng et al. such as a bimodal distribution and a large red enhancement around I = 23 mag. This diagram has not been corrected for the different magnitude systems used by Peng at al. and the HSC, or corrected for extinction. This explains why the mean color for our diagram is V-I ≈ 0.5 (ABMAG), while for Peng et al. it is V-I ≈ 1.1 (VEGAmag system).
  
  We will now show the steps to apply the necessary corrections and re-plot the graphs for a closer inspection.

Part 2: Download and Correct the Data

Part 3: Plot and Compare the Data

• STEP 8- Compare the HSC I filter data.
  
  In this use case we combine the A_F814W and W_F814W magnitudes. We now check whether this step is justified by determining the agreement between the two measurements. This first plot shows the agreement in the raw HSC data; interestingly there appears to be a scattering of points below the main relation. These data are objects on the PC portion of the WFPC2. The problem arises because of the difference in resolution and pixel scale (0.05"/pix for PC and 0.10"/pix for WFC) between the PC and WFC in WFPC2. Hence the two filters require a different aperture correction. These outliers are removed by our CI < 1.05 cut in step 7.
  
  

  
  
  

  
  The figure below (left) shows the result after the sample has been trimmed using steps 6-9. The residual plot (right) shows the scatter in more detail
  
  
  The magnitudes were binned into sections and the average and scatter for each bin were calculated. Based on this comparison we add 0.03 magnitudes to the WFPC2 observations and the value of 23 mag was taken to be the cut off point. The inconsistency is probably due to some Malmquist bias.
  
  


• STEP 9- Match sources and compare photometry.
  
  After applying a small offset of +0.022° RA and +0.009° Dec to the Peng et al. positions we can match the two catalogs. A comparison of the color in ABMAG from both catalogs is shown below. Note the bimodality in the scatter of the points.

  

  
  
  Now we separate the data into high and low quality subsamples. The left figure displays all sources in the matched catalog and the right figure displays only the points with uncertainties <0.1 mag. The dashed line has a slope of unity, whereas the solid lines are linear regression fits to the data points.
  
  


• STEP 10- Create and compare the Color Magnitude Diagrams.
  
  Below is the original Peng et al. CMD for M87 converted into ABMAG magnitudes, plotted alongside the CMDs for M87 constructed using HSC MagAper2 data, and on the far right constructed using HSC MagAuto. Below these diagrams are the corresponding color histograms.

  

  
  Upon inspection the correspondence between Peng et al. and HSC MagAper data is good: both diagrams have a similar range of color and both display a clear bimodality. Peng et al. I magnitudes extend to fainter values because their photometry was gathered from co-added images. The bimodality is strongest for the Peng data as expected since both A_F606W and A_F814W magnitudes were used.
  
  Our previous assumption that, using MagAuto for measuring magnitudes was detrimental, is proven correct in this diagram. Most of the bimodality in this diagram has been washed out.
  
  


• STEP 11- Comparison of the size of the clusters in blue and red enhancements
  
  The blue and red enhancements were separated and the mean CI values were calculated. Below the separation of the two colors is taken as the median.
  
  

  

  The median value of the distribution, 0.513 mag, was taken to be the separation color. The average CI for the blue and red sub-populations were (1.316 ± 0.005) mag and (1.281 ± 0.006) mag respectively, giving a difference between the two of (0.035 ± 0.008) mag. This result indicates a statistically significant size difference between the two filters.
  
  This result, that the radii of blue GCs are larger, is in agreement with Kundu et al. (1999) and also agrees with Webb et al. (2012)
  
  
• STEP 12- Calculate the completeness in magnitude assuming that Peng et al. catalog is "the truth".
  
  Note that ACS and WFPC2 source lists will not have the same coverage. The image (left) is from the footprints tab of the HLA, centered on M87, with A_F606W and W_F606W highlighted. The top left of the ACS square (yellow) has a section that is not included by the WFPC2 coverage and vice versa for other locations of the image. Seen in the image on the right is the image from step 1 with NumImages > 2 and the colors inverted, note the corresponding sky coverage between the two figures.
  Before we can compute the completeness levels, the samples need to be trimmed so we have uniform coverage in both catalogs. In the figure below on the left, we use the same image from hla.stsci.edu from steps 1 & 2, and over-plotted the regions of Peng et al. (green) and HSC (blue). Note the lack of HSC sources in the nuclear region, due to the high background. We impose a maximum outer radius in order to make a fair comparison when computing the completeness, the distribution is shown in the right image.
  
  
  
  The data was sorted into bins of 0.5 mag, and the ratio was taken between each catalog. The green line shows the mean value of all the points before the drop off.
  
  

  

  
  
  The 50% threshold magnitude was calculated using interpolation to be 25.13 mag and 23.98 mag for F606W and F814W filters respectively.
  
  

  

  
  The higher depth of the Peng et al data set is apparent when the total area of the histogram is compared with the HSC histogram. However the two catalogs have similar peaks in I, since the HSC is just deep enough to reach the peak of the luminosity function at I ~ 22.7 (e.g. Kundu et al. (1999))
  

1. The HSC Beta version 0.2 catalog was used to recover the bimodality of the CMD of M87 seen in Peng et al. (2009).
2. The magnitude cutoff was ~2 mag less that Peng et al. (2009) because the photometry of HSC is inferred from single visits rather than a co-added image.
3. Foreground stars can be distinguished from GCs using the concentration index.
4. Aperture magnitudes are the superior method (rather than MagAuto) for calculating the color of clusters.
5. The sizes of the globular clusters in the blue and red sub-populations were calculated for M87 and the difference was (0.035 ± 0.008) mag with the blue being the larger of the two.
6. Relative to the Peng et al. catalog, the HSC is complete to 99% down to ~24 in F606W and ~23 in F814W.

1. Use the HSC to measure the blue tilt of the CMD of M87. We refer the reader to Peng et al. (2009) and Waters et al. (2009) for examples of this topic.
2. Measure the variability of sources (see HSC Use Case #2) in M87 similar to Baltz et al. (2004).

1. Include photometry of more sources by not eliminating those filter magnitudes for sources with only one image, when they have other filter magnitudes with more than one image.